data_source
stringclasses 3
values | prompt
listlengths 2
2
| ability
stringclasses 1
value | reward_model
dict | extra_info
dict |
|---|---|---|---|---|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-4914.csv` in your code execution environment.\n\nTable Title: Watford F.C.\nTable Content:\n| Name | Year inducted | Position | Apps | Goals |\n| --- | --- | --- | --- | --- |\n| Luther Blissett | 2003 | Forward | 503 | 186 |\n| Tony Coton | 2004 | Goalkeeper | 291 | 0 |\n| John McClelland | 2005 | Defender | 234 | 3 |\n| Tommy Mooney | 2006 | Forward | 287 | 64 |\n| Les Taylor | 2007 | Midfielder | 211 | 20 |\n| David James | 2008 | Goalkeeper | 98 | 0 |\n| Ian Bolton | 2009 | Defender | 287 | 36 |\n| Nigel Gibbs | 2010 | Defender | 491 | 7 |\n| Duncan Welbourne | 2011 | Defender | 457 | 25 |\n| Ross Jenkins | 2012 | Forward | 398 | 142 |\n| Robert Page | 2013 | Defender | 218 | 2 |\n\n\nQuestion\nwhat is the total number of apps and goals for luther blissett?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"689"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-4914",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-4200.csv` in your code execution environment.\n\nTable Title: Crazy Town\nTable Content:\n| Title | Year | Peak chart positions\\nUS | Peak chart positions\\nUS\\nAlt. | Peak chart positions\\nUS\\nMain. Rock | Peak chart positions\\nAUS | Peak chart positions\\nAUT | Peak chart positions\\nFIN | Peak chart positions\\nGER | Peak chart positions\\nNLD | Peak chart positions\\nSWI | Peak chart positions\\nUK | Album |\n| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |\n| \"Toxic\" | 1999 | — | — | — | — | — | — | — | — | — | — | The Gift of Game |\n| \"Darkside\" | 2000 | — | — | — | — | — | — | — | — | — | — | The Gift of Game |\n| \"Butterfly\" | 2000 | 1 | 1 | 21 | 4 | 1 | 2 | 1 | 8 | 1 | 3 | The Gift of Game |\n| \"Revolving Door\" | 2001 | — | — | — | 76 | 29 | 19 | 26 | 71 | 43 | 23 | The Gift of Game |\n| \"Drowning\" | 2002 | — | 24 | 24 | — | 45 | — | 45 | — | — | 50 | Darkhorse |\n| \"Hurt You So Bad\" | 2003 | — | — | — | — | — | — | — | — | — | — | Darkhorse |\n\n\nQuestion\nby how many chart positions higher did \"revolving door\" peak in the uk compared to the peak position of \"drowning\" in the uk?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"27"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-4200",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-7094.csv` in your code execution environment.\n\nTable Title: Andy Roddick\nTable Content:\n| Outcome | Year | Championship | Surface | Opponent | Score |\n| --- | --- | --- | --- | --- | --- |\n| Runner-up | 2002 | Canada (Toronto) (1) | Hard | Guillermo Cañas | 4–6, 5–7 |\n| Winner | 2003 | Montreal | Hard | David Nalbandian | 6–1, 6–3 |\n| Winner | 2003 | Cincinnati | Hard | Mardy Fish | 4–6, 7–6(7–3), 7–6(7–4) |\n| Winner | 2004 | Miami | Hard | Guillermo Coria | 6–7(2–7), 6–3, 6–1, ret. |\n| Runner-up | 2004 | Toronto (2) | Hard | Roger Federer | 5–7, 3–6 |\n| Runner-up | 2005 | Cincinnati | Hard | Roger Federer | 3–6, 5–7 |\n| Winner | 2006 | Cincinnati (2) | Hard | Juan Carlos Ferrero | 6–3, 6–4 |\n| Runner-up | 2010 | Indian Wells | Hard | Ivan Ljubičić | 6–7(3–7), 6–7(5–7) |\n| Winner | 2010 | Miami (2) | Hard | Tomáš Berdych | 7–5, 6–4 |\n\n\nQuestion\nhow many times was the championship in miami?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"2"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-7094",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-77.csv` in your code execution environment.\n\nTable Title: MXR Yorkshire\nTable Content:\n| Service | Service ID | Bit rate | Audio channels | Description | Analogue availability |\n| --- | --- | --- | --- | --- | --- |\n| Capital Yorkshire | C3AA | 128 kbit/s | Joint stereo | Dance and R'n'B | 105.1 MHz (South & West Yorkshire)\\n105.6 MHz (Bradford & Sheffield)\\n105.8 MHz (Humberside) |\n| Choice FM | C0CE | 128 kbit/s | Joint stereo | R'n'B and Hip-hop | — |\n| Heart London | CCCD | 128 kbit/s | Joint stereo | Adult contemporary | — |\n| Gold | C2CE | 64 kbit/s | Mono LSF | Oldies, 60s, 70s, 80s | — |\n| Real Radio (Yorkshire) | C3A4 | 128 kbit/s | Joint stereo | Contemporary | 106.2 MHz (South & West Yorkshire)\\n107.6 MHz (Bradford & Halifax)\\n107.7 MHz (Sheffield) |\n| Real Radio XS | C8C5 | 128 kbit/s | Joint stereo | Classic rock | — |\n| UCB Gospel | CED0 | 64 kbit/s | Mono LSF | Christian | — |\n| UCB Inspirational | CFD0 | 64 kbit/s | Mono LSF | Christian | — |\n| XFM | C0B8 | 128 kbit/s | Joint stereo | Indie / alternative / rock music | — |\n| Yorkshire Radio | CFCD | 128 kbit/s | Joint stereo | Pop music and sport | — |\n| DAB Guide | E1C000A9 | 8 kbit/s | Data | Hidden service, EPG data | — |\n\n\nQuestion\nwhat is the number of christian radio stations broadcasted by mxr yorkshire?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"2"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-77",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-12591.csv` in your code execution environment.\n\nTable Title: Luís Jesus\nTable Content:\n| Year | Competition | Venue | Position | Event | Notes |\n| --- | --- | --- | --- | --- | --- |\n| 1998 | Turin Marathon | Turin, Italy | 3rd | Marathon | 2:12:59 |\n| 2003 | Lisbon Marathon | Lisbon, Portugal | 1st | Marathon | 2:15:31 |\n| 2004 | Vienna Marathon | Vienna, Austria | 3rd | Marathon | 2:11:24 |\n| 2004 | Berlin Marathon | Berlin, Germany | 5th | Marathon | 2:09:08 |\n| 2005 | World Championships | Helsinki, Finland | 18th | Marathon | 2:16:33 |\n| 2006 | Paris Marathon | Paris, France | 4th | Marathon | 2:08:55 |\n| 2006 | Lisbon Marathon | Lisbon, Portugal | 1st | Marathon | 2:21:08 |\n| 2007 | World Championships | Osaka, Japan | — | Marathon | DNS |\n\n\nQuestion\nhow long did luis jesus take to finish in turin,italy in 1998?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"2:12:59"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-12591",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-9444.csv` in your code execution environment.\n\nTable Title: 2012–13 UCI America Tour\nTable Content:\n| Date | Race Name | Location | UCI Rating | Winner | Team |\n| --- | --- | --- | --- | --- | --- |\n| 6 January | Copa América de Ciclismo | Brazil | 1.2 | Francisco Chamorro (ARG) | Funvic Brasilinvest-São José dos Campos |\n| 11–20 January | Vuelta al Táchira | Venezuela | 2.2 | Yeison Delgado (VEN) | Kino Táchira-Drodínica |\n| 21–27 January | Tour de San Luis | Argentina | 2.1 | Daniel Díaz (ARG) | San Luis Somos Todos |\n| 21–28 April | Vuelta a Guatemala | Guatemala | 2.2 | Óscar Sánchez (COL) | GW Shimano |\n| 1–5 May | Tour of the Gila | United States | 2.2 | Philip Deignan (IRL) | UnitedHealthcare |\n| 2 May | Pan American Cycling Championships – Time Trial | Mexico | CC | Carlos Oyarzun (CHI) | Chile national team |\n| 5 May | Pan American Cycling Championships – Road Race | Mexico | CC | Jonathan Paredes (COL) | Colombia national team |\n| 12–19 May | Tour of California | United States | 2.HC | Tejay Van Garderen (USA) | BMC Racing Team |\n| 2 June | The Philly Cycling Classic | United States | 1.2 | Kiel Reijnen (USA) | UnitedHealthcare |\n| 7–9 June | Coupe des Nations Ville Saguenay | Canada | 2.Ncup | Sondre Enger (NOR) | Norway national team |\n| 9–23 June | Vuelta a Colombia | Colombia | 2.2 | Óscar Sevilla (ESP) | EPM-UNE |\n| 11–16 June | Tour de Beauce | Canada | 2.2 | Nathan Brown (USA) | Bontrager Cycling Team |\n| 7 July | Tour de Delta | Canada | 1.2 | Steve Fisher (USA) | Hagens Berman Cycling |\n| 19–28 July | Vuelta a Venezuela | Venezuela | 2.2 | Carlos José Ochoa (VEN) | Androni Giocattoli-Venezuela |\n| 2–4 August | Tour of Elk Grove | United States | 2.1 | Elia Viviani (ITA) | Cannondale |\n| 6–11 August | Tour of Utah | United States | 2.1 | Tom Danielson (USA) | Garmin-Sharp |\n| 6–11 August | Ruta del Centro | Mexico | 2.2 | Víctor García (ESP) | Depredadores PBG Design |\n| 14–18 August | Vuelta al Sur de Bolivia | Bolivia | 2.2 | Óscar Soliz (BOL) | Pio Rico |\n| 19–25 August | USA Pro Cycling Challenge | United States | 2.HC | Tejay Van Garderen (USA) | BMC Racing Team |\n| 28 Aug–1 Sep | Tour do Rio | Brazil | 2.2 | Óscar Sevilla (ESP) | EPM-UNE |\n| 3–8 September | Tour of Alberta | Canada | 2.1 | Rohan Dennis (AUS) | Garmin-Sharp |\n| 7 September | Bucks County Classic | United States | 1.2 | Kiel Reijnen (USA) | UnitedHealthcare |\n\n\nQuestion\nwhat is the winner name and team for the copa américa de ciclismo race?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"Francisco Chamorro (ARG)",
"Funvic Brasilinvest-São José dos Campos"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-9444",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-12634.csv` in your code execution environment.\n\nTable Title: Mexican Grand Prix\nTable Content:\n| Year | Driver | Constructor | Location | Report |\n| --- | --- | --- | --- | --- |\n| 1992 | Nigel Mansell | Williams-Renault | Hermanos Rodriguez | Report |\n| 1991 | Riccardo Patrese | Williams-Renault | Hermanos Rodriguez | Report |\n| 1990 | Alain Prost | Ferrari | Hermanos Rodriguez | Report |\n| 1989 | Ayrton Senna | McLaren-Honda | Hermanos Rodriguez | Report |\n| 1988 | Alain Prost | McLaren-Honda | Hermanos Rodriguez | Report |\n| 1987 | Nigel Mansell | Williams-Honda | Hermanos Rodriguez | Report |\n| 1986 | Gerhard Berger | Benetton-BMW | Hermanos Rodriguez | Report |\n| 1985\\n-\\n1971 | Not held | Not held | Not held | Not held |\n| 1970 | Jacky Ickx | Ferrari | Magdalena Mixhuca | Report |\n| 1969 | Denny Hulme | McLaren-Ford | Magdalena Mixhuca | Report |\n| 1968 | Graham Hill | Lotus-Ford | Magdalena Mixhuca | Report |\n| 1967 | Jim Clark | Lotus-Ford | Magdalena Mixhuca | Report |\n| 1966 | John Surtees | Cooper-Maserati | Magdalena Mixhuca | Report |\n| 1965 | Richie Ginther | Honda | Magdalena Mixhuca | Report |\n| 1964 | Dan Gurney | Brabham-Climax | Magdalena Mixhuca | Report |\n| 1963 | Jim Clark | Lotus-Climax | Magdalena Mixhuca | Report |\n| 1962 | Trevor Taylor\\n Jim Clark | Lotus-Climax | Magdalena Mixhuca | Report |\n\n\nQuestion\nwho won the year before alain prost?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"Riccardo Patrese"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-12634",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-160.csv` in your code execution environment.\n\nTable Title: France at the 1932 Summer Olympics\nTable Content:\n| Medal | Name | Sport | Event | Date |\n| --- | --- | --- | --- | --- |\n| Gold | Louis Chaillot, Maurice Perrin | Cycling | Men's tandem | August 3 |\n| Gold | Xavier Lesage | Equestrian | Individual dressage | August 10 |\n| Gold | André Jousseaume, Xavier Lesage,\\nCharles Marion | Equestrian | Team dressage | August 10 |\n| Gold | Georges Buchard, Philippe Cattiau,\\nFernand Jourdant, Jean Piot,\\nBernard Schmetz, Georges Tainturier | Fencing | Men's team épée | August 7 |\n| Gold | René Bondoux, René Bougnol,\\nPhilippe Cattiau, Edward Gardère,\\nRené Lemoine, Jean Piot | Fencing | Men's team foil | August 1 |\n| Gold | Jacques Lebrun | Sailing | Snowbird class | August 12 |\n| Gold | Raymond Suvigny | Weightlifting | Men's 60 kg | July 31 |\n| Gold | René Duverger | Weightlifting | Men's 67.5 kg | July 30 |\n| Gold | Louis Hostin | Weightlifting | Men's 82.5 kg | July 30 |\n| Gold | Charles Pacôme | Wrestling | Men's freestyle lightweight | August 3 |\n| Silver | Louis Chaillot | Cycling | Men's sprint | August 3 |\n| Silver | Paul Chocque, Amédée Fournier,\\nRené Le Grèves, Henri Mouillefarine | Cycling | Men's team pursuit | August 2 |\n| Silver | Charles Marion | Equestrian | Individual dressage | August 10 |\n| Silver | Georges Buchard | Fencing | Men's épée | August 9 |\n| Silver | Jean Taris | Swimming | Men's 400 m freestyle | August 10 |\n| Bronze | Paul Winter | Athletics | Men's discus throw | August 3 |\n| Bronze | Charles Rampelberg | Cycling | Men's 1000 m time trial | August 1 |\n| Bronze | Pierre Brunet, Anselme Brusa,\\nAndré Giriat | Rowing | Men's coxed pair | August 13 |\n| Bronze | Louis François | Wrestling | Men's Greco-Roman bantamweight | August 7 |\n\n\nQuestion\nhow many gold medals did this country win during these olympics?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"10"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-160",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-13551.csv` in your code execution environment.\n\nTable Title: Slipping Away (Moby song)\nTable Content:\n| Version | Length | Album | Remixed by | Year | Comment |\n| --- | --- | --- | --- | --- | --- |\n| Single Version / Radio Edit | 3:39 | Go – The Very Best of Moby | MHC aka Philip Larsen and Chis Smith | 2006 | Remixed version with dance sonorities, with English and French lyrics which alternate. This version is more rhythmic than the original one. |\n| Extended Remix | 6:48 | — | MHC aka Philip Larsen and Chis Smith | 2006 | Similar to the previous but longer. |\n| Axwell Remix | 7:24 | — | Axwell | 2006 | Used in a scene in the 2008 movie 21. |\n| Enzo Mori & Stephan Clark Remix | 7:10 | — | Enzo Mori and Stephan Clark | 2006 | |\n| MHC club Remix | 7:31 | — | MHC aka Philip Larsen and Chis Smith | 2006 | |\n| Zloot Remix | 4:36 | — | Toni Toolz | 2006 | |\n| Music Video | 3:58 | — | — | 2006 | |\n\n\nQuestion\nwhat is the total amount of remix versions?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"5"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-13551",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-5722.csv` in your code execution environment.\n\nTable Title: Biathlon at the Winter Paralympics\nTable Content:\n| Rank | Nation | Gold | Silver | Bronze | Total |\n| --- | --- | --- | --- | --- | --- |\n| 1 | Germany | 17 | 9 | 14 | 40 |\n| 2 | Russia | 12 | 16 | 8 | 36 |\n| 3 | Ukraine | 6 | 11 | 9 | 26 |\n| 4 | Norway | 6 | 2 | 7 | 15 |\n| 5 | France | 5 | 4 | 3 | 12 |\n| 6 | Switzerland | 3 | 5 | 3 | 11 |\n| 7 | Finland | 3 | 2 | 4 | 9 |\n| 8 | Japan | 2 | 2 | 1 | 5 |\n| 9 | Netherlands | 2 | 0 | 1 | 3 |\n| 10 | Sweden | 1 | 2 | 2 | 5 |\n| 11 | Denmark | 1 | 0 | 1 | 2 |\n| 12 | Unified Team | 1 | 0 | 0 | 1 |\n| 13 | Austria | 0 | 4 | 0 | 4 |\n| 14 | Slovakia | 0 | 2 | 1 | 3 |\n| 15 | Italy | 0 | 0 | 1 | 1 |\n| 15 | Poland | 0 | 0 | 1 | 1 |\n| 15 | Canada | 0 | 0 | 1 | 1 |\n| 15 | Belarus | 0 | 0 | 1 | 1 |\n| 15 | United States | 0 | 0 | 1 | 1 |\n| Total 19 nations | Total 19 nations | 59 | 59 | 59 | 177 |\n\n\nQuestion\nonly one team has won more than 10 bronze medals in the biathlon at the winter paralympics. which team is it?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"Germany"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-5722",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-4481.csv` in your code execution environment.\n\nTable Title: List of lakes of Turkey\nTable Content:\n| Name in English | Name in Turkish | Area (km2) | Depth | Location (districts and/or provinces) |\n| --- | --- | --- | --- | --- |\n| Lake Van | Van Gölü | 3755 km2 | 451 m | Van, Bitlis |\n| Lake Tuz | Tuz Gölü | 1500 km2 | 2 m | Aksaray, Ankara, Konya |\n| Lake Beyşehir | Beyşehir Gölü | 656 km2 | 10 m | Beyşehir in Konya, Isparta |\n| Lake Eğirdir | Eğirdir Gölü | 482 km2 | | Isparta |\n| Lake İznik | İznik Gölü | 308 km2 | | İznik in Bursa, Yalova |\n| Lake Burdur | Burdur Gölü | 200 km2 | | Burdur, Isparta |\n| Lake Manyas | Manyas Gölü | 166 km2 | | Balıkesir |\n| Lake Acıgöl | Acıgöl | 153 km2 | | Denizli, Afyonkarahisar |\n| Lake Uluabat | Uluabat Gölü | 134 km2 | 1–2 m | Bursa |\n| Lake Çıldır | Çıldır Gölü | 115 km2 | | Ardahan, Kars |\n| Lake Palas Tuzla | Palas Tuzla Gölü | 106 km2 | 15 m | Palas/Kayseri |\n| Lake Akşehir | Akşehir Gölü | 105 km2 | | Akşehir in Konya, Afyonkarahisar |\n| Lake Eber | Eber Gölü | 104 km2 | | Afyonkarahisar |\n| Lake Erçek | Erçek Gölü | 98 km2 | | Van |\n| Lake Hazar | Hazar Gölü | 86 km2 | | Elazığ |\n| Lake Bafa | Bafa Gölü | 60 km2 | | Aydın, Muğla |\n| Lake Köyceğiz | Köyceğiz Gölü | 52 km2 | | Köyceğiz in Muğla |\n| Lake Işıklı | Işıklı Gölü | 49 km2 | | Denizli |\n| Lake Nazik | Nazik Gölü | 48 km2 | | Bitlis |\n| Lake Sapanca | Sapanca Gölü | 47 km2 | | Sakarya Province |\n| Lake Salda | Salda Gölü | 45 km2 | 184 m | Burdur |\n| Lake Yay | Yay Gölü | 37 km2 | | Kayseri |\n| Lake Akyatan | Akyatan Gölü | 35 km2 | | Adana |\n| Lake Balık | Balık Gölü | 34 km2 | | Doğubeyazıt in Ağrı |\n| Lake Marmara | Marmara Gölü | 34 km2 | | Salihli, Gölmarmara in Manisa |\n| Lake Çöl | Çöl Gölü | 32 km2 | | Ankara |\n| Lake Durusu (Lake Terkos) | Durusu Gölü | 25 km2 | | İstanbul |\n| Lake Karine | Karine Gölü | 24 km2 | | |\n| Lake Tuzla | Tuzla Gölü | 23 km2 | | Tuzla |\n| Lake Küçükçekmece | Küçükçekmece Gölü | 16 km2 | | Küçükçekmece, İstanbul |\n| Lake Yaraşlı | Yaraşlı Gölü | 16 km2 | | Burdur |\n| Lake Haçlı | Haçlı Gölü | 16 km2 | | Muş |\n| Lake Seyfe | Seyfe Gölü | 15 km2 | | Kırşehir |\n| Lake Akyayan | Akyayan Gölü | 15 km2 | | |\n| Lake Hozapin | Hozapin Gölü | 14 km2 | | |\n| Lake Arin | Arin Gölü | 13 km2 | | |\n| Lake Nemrut | Nemrut Gölü | 12 km2 | | Bitlis Province |\n| Lake Balık | Balık Gölü | 12 km2 | | |\n| Lake Büyükçekmece | Büyükçekmece Gölü | 11 km2 | | Büyükçekmece, Istanbul |\n| Lake Boluk | Boluk Gölü | 11 km2 | | |\n| Lake Akdoğan | Akdoğan Gölü | 11 km2 | | |\n| Lake Çavuşlu | Çavuşlu Gölü | 9 km2 | | |\n| Lake Düden | Düden Gölü | 8 km2 | | |\n| Lake Gala | Gala Gölü | 8 km2 | | Edirne |\n| Lake Karataş | Karataş Gölü | 6 km2 | | |\n| Lake Mogan | Mogan Gölü | 6 km2 | | Ankara |\n| Paradeniz | Paradeniz | 4 km2 | | Mersin |\n| Lake Eymir | Eymir Gölü | 1.8 km2 | | Ankara |\n| Lake Abant | Abant Gölü | 1.28 km2 | 18 m | Bolu |\n| Lake Gölcük | Gölcük Gölü | 1 km2 | | İzmir |\n\n\nQuestion\nwhich lake has the most area and it locations\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"Lake Van",
"Van, Bitlis"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-4481",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-13885.csv` in your code execution environment.\n\nTable Title: Zbigniew Rybczyński\nTable Content:\n| Year | Film Title | Duration | Film Type | Shoot Location |\n| --- | --- | --- | --- | --- |\n| 1972 | Kwadrat (Square) | 4:40 | 35 mm short film | PWSFTviT Łódź, Poland |\n| 1972 | Take Five | 3:36 | 35 mm short film | PWSFTviT Łódź, Poland |\n| 1973 | Plamuz (Music Art) | 9:38 | 35 mm short film | SMFF Se-Ma-For Łódź, Poland |\n| 1974 | Zupa (Soup) | 8:22 | 35 mm short film | SMFF Se-Ma-For Łódź, Poland |\n| 1975 | Nowa Książka (New Book) | 10:26 | 35 mm short film | SMFF Se-Ma-For Łódź, Poland |\n| 1975 | Lokomotywa (Locomotive) | 9:38 | 35 mm short film | SMFF Se-Ma-For Łódź, Poland |\n| 1975 | Święto (Holiday) | 9:38 | 35 mm short film | SMFF Se-Ma-For Łódź, Poland |\n| 1976 | Oj! Nie Mogę Się Zatrzymać! (Oh, I Can't Stop!) | 10:07 | 35 mm short film | SMFF Se-Ma-For Łódź, Poland |\n| 1976 | Weg Zum Nachbarn (Way To Your Neighbor) | 2:30 | 35 mm short film | SMFF Se-Ma-For Łódź, Poland |\n| 1977 | Piątek - Sobota (Friday - Saturday) | 3:00 | 35 mm short film | SMFF Se-Ma-For Łódź, Poland |\n| 1979 | Mein Fenster (My Window) | 2:26 | 35 mm short film | Vienna |\n| 1980 | Tango | 8:14 | 35 mm short film | SMFF Se-Ma-For Łódź, Poland |\n| 1980 | Media | 1:36 | 35 mm short film | SMFF Se-Ma-For Łódź, Poland |\n| 1980 | Sceny Narciarskie z Franzem Klammeren (Ski Scenes with Franz Klammer) | 9:38 | 35 mm documentary film | in collaboration with B. Dziworski, WFO Łódź, Poland and Signal Film, Vienna |\n| 1981 | Wdech-Wydech (Inhale-Exhale) | 2:26 | 35 mm short film | in collaboration with B. Dziworski, SMFF Se-Ma-For Łódź, for TVP, Poland |\n\n\nQuestion\nhow many films had a duration below 5 minutes?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"7"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-13885",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-415.csv` in your code execution environment.\n\nTable Title: United States presidential election in New York, 2004\nTable Content:\n| County | Kerry | Votes | Bush | Votes | Others | Votes |\n| --- | --- | --- | --- | --- | --- | --- |\n| Bronx | 82.8% | 283,994 | 16.5% | 56,701 | 0.7% | 2,284 |\n| Manhattan | 82.1% | 526,765 | 16.7% | 107,405 | 1.2% | 7,781 |\n| Brooklyn | 74.9% | 514,973 | 24.3% | 167,149 | 0.8% | 5,762 |\n| Queens | 71.7% | 433,835 | 27.4% | 165,954 | 0.9% | 5,603 |\n| Tompkins | 64.2% | 27,229 | 33.0% | 13,994 | 2.8% | 1,179 |\n| Albany | 60.7% | 89,323 | 37.3% | 54,872 | 2.0% | 3,004 |\n| Westchester | 58.1% | 229,849 | 40.3% | 159,628 | 1.6% | 6,293 |\n| Erie | 56.4% | 251,090 | 41.4% | 184,423 | 2.2% | 9,625 |\n| Saint Lawrence | 54.7% | 22,857 | 43.2% | 18,029 | 2.1% | 869 |\n| Ulster | 54.3% | 47,602 | 43.1% | 37,821 | 2.6% | 2,289 |\n| Onondaga | 54.2% | 116,381 | 43.8% | 94,006 | 2.0% | 4,202 |\n| Nassau | 52.2% | 323,070 | 46.6% | 288,355 | 1.1% | 6,918 |\n| Clinton | 52.2% | 17, 624 | 45.4% | 15,330 | 2.3% | 782 |\n| Franklin | 52.1% | 9,543 | 45.8% | 8,383 | 2.1% | 390 |\n| Schenectady | 51.8% | 35,971 | 46.2% | 32,066 | 2.1% | 1,432 |\n| Columbia | 51.2% | 15,929 | 46.5% | 14,457 | 2.3% | 717 |\n| Monroe | 50.6% | 173,497 | 47.7% | 163,545 | 1.7% | 5,939 |\n| Broome | 50.4% | 46,281 | 47.4% | 43,568 | 2.2% | 2,041 |\n| Rensselaer | 49.7% | 36,075 | 47.9% | 34,734 | 2.4% | 1,705 |\n| Suffolk | 49.5% | 315,909 | 48.5% | 309,949 | 2.0% | 12,854 |\n| Niagara | 49.3% | 47,602 | 48.8% | 47,111 | 1.9% | 1,867 |\n| Rockland | 48.9% | 64,191 | 49.6% | 65,130 | 1.5% | 1,910 |\n| Cayuga | 48.6% | 17,534 | 49.2% | 17,743 | 2.1% | 775 |\n| Sullivan | 48.6% | 15,034 | 49.5% | 15,319 | 2.0% | 613 |\n| Otsego | 47.7% | 12,723 | 50.1% | 13,342 | 2.2% | 587 |\n| Dutchess | 47.0% | 58,232 | 51.2% | 63,372 | 1.8% | 2,277 |\n| Cortland | 46.9% | 10,670 | 51.0% | 11,613 | 2.1% | 477 |\n| Oswego | 46.8% | 24,133 | 51.0% | 26,325 | 2.2% | 1,149 |\n| Essex | 45.9% | 8,768 | 51.7% | 9,869 | 2.3% | 445 |\n| Saratoga | 45.6% | 48,730 | 52.5% | 56,158 | 1.9% | 1,985 |\n| Seneca | 45.5% | 6,979 | 52.1% | 7,981 | 2.4% | 365 |\n| Chautauqua | 44.7% | 27,257 | 53.2% | 32,434 | 2.1% | 1,251 |\n| Montgomery | 44.5% | 9,449 | 53.4% | 11,338 | 2.0% | 434 |\n| Orange | 43.8% | 63,394 | 54.7% | 79,089 | 1.5% | 2,190 |\n| Chemung | 43.7% | 17,080 | 54.6% | 21,321 | 1.7% | 674 |\n| Chenango | 43.5% | 9,277 | 54.3% | 11,582 | 2.3% | 482 |\n| Jefferson | 43.5% | 16,860 | 54.7% | 21,231 | 1.8% | 709 |\n| Madison | 43.3% | 13,121 | 54.6% | 16,537 | 2.1% | 629 |\n| Warren | 43.2%' | 13,405 | 54.6% | 16,969 | 2.2% | 685 |\n| Oneida | 42.8% | 40,792 | 54.9% | 52,392 | 2.3% | 2,185 |\n| Richmond (Staten Island) | 42.7% | 68,448 | 56.4% | 90,325 | 0.9% | 1,370 |\n| Washington | 42.3% | 10,624 | 55.1% | 13,827 | 2.6% | 652 |\n| Ontario | 42.2% | 21,166 | 55.9% | 27,999 | 1.9% | 937 |\n| Putnam | 42.0% | 19,575 | 56.6% | 26,356 | 1.4% | 632 |\n| Fulton | 41.4% | 9,202 | 56.6% | 12,570 | 2.0% | 443 |\n| Delaware | 41.2% | 8,724 | 56.5% | 11,958 | 2.3% | 484 |\n| Herkimer | 41.2% | 11,675 | 56.6% | 16,024 | 2.2% | 611 |\n| Tioga | 40.6% | 9,694 | 57.6% | 13,762 | 1.9% | 446 |\n| Schuyler | 40.1% | 3,445 | 57.7% | 4,960 | 2.2% | 185 |\n| Greene | 39.9% | 8,933 | 58.0% | 12,996 | 2.1% | 469 |\n| Lewis | 39.9% | 4,546 | 58.1% | 6,624 | 2.0% | 227 |\n| Cattaraugus | 39.4% | 13,514 | 58.5% | 20,051 | 2.0% | 697 |\n| Yates | 39.3% | 4,205 | 58.9% | 6,309 | 1.8% | 197 |\n| Schoharie | 38.7% | 5,630 | 59.0% | 8,591 | 2.3% | 336 |\n| Livingston | 38.4% | 11,504 | 59.2% | 17,729 | 2.4% | 715 |\n| Wayne | 38.1% | 15,709 | 60.0% | 24,709 | 1.9% | 782 |\n| Genesee | 37.5% | 10,331 | 60.6% | 16,725 | 1.9% | 524 |\n| Orleans | 36.0% | 5,959 | 62.3% | 10,317 | 1.8% | 297 |\n| Steuben | 34.3% | 14,523 | 63.8% | 26,980 | 1.8% | 781 |\n| Allegany | 34.1% | 6,566 | 63.9% | 12,310 | 2.0% | 394 |\n| Wyoming | 33.8% | 6,134 | 64.7% | 11,745 | 1.6% | 285 |\n| Hamilton | 31.0% | 1,145 | 67.0% | 2,475 | 2.0% | 72 |\n\n\nQuestion\nwhat is the different of votes between both in orange county\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"15695"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-415",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-4196.csv` in your code execution environment.\n\nTable Title: Kathrin Hölzl\nTable Content:\n| Season | Date | Location | Discipline | Place |\n| --- | --- | --- | --- | --- |\n| 2007 | 18 Mar 2007 | Lenzerheide, Switzerland | Giant Slalom | 2nd |\n| 2009 | 10 Jan 2009 | Maribor, Slovenia | Giant Slalom | 3rd |\n| 2010 | 28 Nov 2009 | Aspen, USA | Giant Slalom | 1st |\n| 2010 | 28 Dec 2009 | Lienz, Austria | Giant Slalom | 1st |\n| 2010 | 24 Jan 2010 | Cortina d'Ampezzo, Italy | Giant Slalom | 3rd |\n| 2010 | 11 Mar 2010 | Garmisch, Germany | Giant Slalom | 2nd |\n| 2011 | 23 Oct 2010 | Sölden, Austria | Giant Slalom | 2nd |\n| 2011 | 27 Nov 2010 | Aspen, USA | Giant Slalom | 3rd |\n| 2011 | 28 Dec 2010 | Semmering, Austria | Giant Slalom | 3rd |\n\n\nQuestion\nwhat is the total number of her 2nd place finishes on the list?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"3"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-4196",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-554.csv` in your code execution environment.\n\nTable Title: Power-to-weight ratio\nTable Content:\n| Capacitor type | Capacity | Volts | Temp. | Energy-to-weight ratio | Power-to-weight ratio |\n| --- | --- | --- | --- | --- | --- |\n| ACT Premlis Lithium ion capacitor | 2000 F | 4.0 V | 25 °C | 54 kJ/kg to 2.0 V | 44.4 W/kg @ 5 A |\n| ACT Premlis Lithium ion capacitor | 2000 F | 4.0 V | 25 °C | 31 kJ/kg to 2.0 V | 850 W/kg @ 10 A |\n| Nesccap Electric double-layer capacitor | 5000 F | 2.7 V | 25 °C | 19.58 kJ/kg to 1.35 V | 5.44 W/kg C/1 (1.875 A) |\n| Nesccap Electric double-layer capacitor | 5000 F | 2.7 V | 25 °C | 5.2 kJ/kg to 1.35 V | 5,200 W/kg @ 2,547A |\n| EEStor EESU barium titanate supercapacitor | 30.693 F | 3500 V | 85 °C | 1471.98 kJ/kg | 80.35 W/kg C/5 |\n| EEStor EESU barium titanate supercapacitor | 30.693 F | 3500 V | 85 °C | 1471.98 kJ/kg | 8,035 W∕kg 20 C |\n| General Atomics 3330CMX2205 High Voltage Capacitor | 20.5 mF | 3300 V | ? °C | 2.3 kJ/kg | 6.8 MW/kg @ 100 kA |\n\n\nQuestion\nwhich one is previous from eesu\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"Nesccap Electric double-layer capacitor"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-554",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-9576.csv` in your code execution environment.\n\nTable Title: Bhanumathi Ramakrishna\nTable Content:\n| Year | Film | Language | Character | Co-Star | Music | Banner |\n| --- | --- | --- | --- | --- | --- | --- |\n| 1939 | Vara Vikrayam | Telugu | Kalindi | Chittor V. Nagaiah | Prabhala Satyanarayana | East India Film Company |\n| 1940 | Malathi Madhavam | Telugu | | | | East India Film Company |\n| 1941 | Bhaktimala | Telugu | Radha | | | Bhaskar Films |\n| 1941 | Dharmapatni | Tamil | | | | Famous Films |\n| 1941 | Dharmapatni | Telugu | | | | Famous Films |\n| 1943 | Garuda Garvabhangam | Telugu | | | | Pratiba |\n| 1943 | Krishna Prema | Telugu | Chandravali | | Galipenchala Narasimha Rao | Famous Films & Star Combines |\n| 1944 | Tahsildar | Telugu | Kamala | | | |\n| 1945 | Swargaseema | Telugu | Subbi/Sujatha Devi | Chittor V. Nagaiah | Chittor V. Nagaiah | Vijaya Studios |\n| 1946 | Gruhapravesam | Telugu | Janaki | | | Sarati Films |\n| 1947 | Ratnamala | Telugu | Ratnamala | C. S. R. Anjaneyulu | C. R. Subburaman | Bharani Pictures |\n| 1948 | Raja Mukthi | Tamil | Kannika | M. K. Thyagaraja Bhagavathar | C. R. Subburaman | Narendra Pictures |\n| 1949 | Laila Majnu | Telugu | Laila | A. Nageswara Rao | C. R. Subburaman | Bharani Pictures |\n| 1949 | Laila Majnu | Tamil | Laila | A. Nageswara Rao | C. R. Subburaman | Bharani Pictures |\n| 1949 | Nallathambi | Tamil | Pushpa | N. S. Krishnan | C. R. Subburaman | NSK Films & Uma Pictures |\n| 1949 | Raksharekha | Telugu | Princess Kalavathi | A. Nageswara Rao | Ogirala Ramachandra Rao & H. R. Padmanabha Sastri | |\n| 1949 | Devamanohari | Tamil | Devamanohari | C. Honnappa Bhagavathar | G. Ramanathan | Arkays Pictures |\n| 1949 | Apoorva Sagodharargal | Tamil | Kanchana | M. K. Radha | S. Rajeswara Rao | Gemini Studios |\n| 1949 | Apoorva Sahodaralu | Telugu | Ranjana | M. K. Radha | S. Rajeswara Rao | Gemini Studios |\n| 1949 | Nishaan | Hindi | Kanchana | M. K. Radha | S. Rajeswara Rao | Gemini Studios |\n| 1949 | Rathnakumar | Tamil | | P. U. Chinnappa Bhagavathar | G. Ramanathan & C. R. Subburaman | Murugan Talkies |\n| 1950 | Maya Rambha | Telugu | | | | |\n| 1951 | Malleeswari | Telugu | Malleeswari | N. T. Rama Rao | Adepalli Rama Rao & S. Rajeswara Rao | Vijaya Studios |\n| 1951 | Mangala | Telugu | Mangala | | | Gemini Studios |\n| 1951 | Mangala | Hindi | Mangala | | | Gemini Studios |\n| 1952 | Prema | Telugu | Mothi | A. Nageswara Rao | C. R. Subburaman | Bharani Pictures |\n| 1952 | Kadhal | Tamil | Mothi | A. Nageswara Rao | C. R. Subburaman | Bharani Pictures |\n| 1952 | Rani | Tamil | | S. Balachander | C. R. Subburaman | Jupiter Pictures |\n| 1952 | Rani | Hindi | | Anoop Kumar | C. R. Subburaman | Jupiter Pictures |\n| 1953 | Chandirani | Tamil | Champa & Chandi | N. T. Rama Rao | C. R. Subburaman, Assistants: M. S. Viswanathan & T. K. Ramamoorthy | Bharani Pictures |\n| 1953 | Chandirani | Telugu | Champa & Chandi | N. T. Rama Rao | C. R. Subburaman, Assistants: M. S. Viswanathan & T. K. Ramamoorthy | Bharani Pictures |\n| 1953 | Chandirani | Hindi | Champa & Chandi | Dilip Kumar | C. R. Subburaman, Assistants: M. S. Viswanathan & T. K. Ramamoorthy | Bharani Pictures |\n| 1954 | Aggi Ramudu | Telugu | Saradha | N. T. Rama Rao | S. M. Subbaiah Naidu | Pakshiraja Studios |\n| 1954 | Malaikkallan | Tamil | Poonkothai | M. G. Ramachandran | S. M. Subbaiah Naidu | Pakshiraja Studios |\n| 1954 | Chakrapani | Telugu | Granddaughter | A. Nageswara Rao | P. Bhanumathi | |\n| 1955 | Vipranarayana | Telugu | Devadevi | A. Nageswara Rao | S. Rajeswara Rao | Bharani Pictures |\n| 1955 | Vipranarayana | Tamil | Devadevi | A. Nageswara Rao | S. Rajeswara Rao | Bharani Pictures |\n| 1955 | Kalvanin Kadhali | Tamil | Kalyani | Sivaji Ganesan | G. Govindarajulu Naidu & Ghantasala | Revathi Productions |\n| 1956 | Alibabavum Narpadhu Thirudargalum | Tamil | Marjina | M. G. Ramachandran | S. Dakshinamurthy | Modern Theatres |\n| 1956 | Alibaba 40 Dongalu | Telugu | Marjina | M. G. Ramachandran | S. Dakshinamurthy | Modern Theatres |\n| 1956 | Chintamani | Telugu | Chintamani | | | |\n| 1956 | Madurai Veeran | Tamil | Princess Bommi | M. G. Ramachandran | G. Ramanathan | Krishna Pictures |\n| 1956 | Rangoon Radha | Tamil | Rangam | Sivaji Ganesan | T. R. Papa | Mekala Pictures |\n| 1956 | Tenali Ramakrishna | Telugu | Rangasaani | N. T. Rama Rao & A. Nageswara Rao | Viswanathan Ramamoorthy | Vikram Productions |\n| 1956 | Tenali Raman | Tamil | Krishnasani | N. T. Rama Rao & Sivaji Ganesan | Viswanathan Ramamoorthy | Vikram Productions |\n| 1956 | Sadharam | Tamil | Sadharam | Gemini Ganesan & K. R. Ramaswamy | G. Ramanathan | Kasturi Films |\n| 1956 | Thaaikkuppin Thaaram | Tamil | Sivakami | M. G. Ramachandran | K. V. Mahadevan | Devar Films |\n| 1956 | Rambaiyin Kaadhal | Tamil | Ramba | K. A. Thangavelu | T. R. Papa | Kalpana Pictures |\n| 1957 | Makkalai Petra Magarasi | Tamil | Rangamma | Sivaji Ganesan | K. V. Mahadevan | Sri Lakshmi Pictures |\n| 1957 | Manamagan Thevai | Tamil | | Sivaji Ganesan | G. Ramanathan | Bharani Pictures |\n| 1957 | Varudu Kavali | Telugu | | K. Jaggaiah | G. Ramanathan | Bharani Pictures |\n| 1957 | Rani Lalithangi | Tamil | | Sivaji Ganesan | G. Ramanathan | TNR Productions |\n| 1957 | Ambikapathy | Tamil | Princess Amaravati | Sivaji Ganesan | G. Ramanathan | ALS Productions |\n| 1957 | Nala Damayanti | Telugu | Damayanthi | Kemparaj Urs | | |\n| 1957 | Sarangadhara | Telugu | Chitrangi Devi | N. T. Rama Rao | G. Ramanathan | Minerva Pictures |\n| 1958 | Sarangadhara | Tamil | Chitrangi Devi | Sivaji Ganesan | G. Ramanathan | Minerva Pictures |\n| 1958 | Nadodi Mannan | Tamil | Madhana | M. G. Ramachandran | S. M. Subbaiah Naidu | Emgeeyar Pictures |\n| 1959 | Manimekalai | Tamil | Manimekalai | T. R. Mahalingam | G. Ramanathan | Shekhar Art Film |\n| 1960 | Raja Bakthi | Tamil | Maharasai | Sivaji Ganesan, Vyjayanthimala | G. Govindarajulu Naidu | Revathi Productions |\n| 1960 | Raja Desingu | Tamil | Ranibhai | M. G. Ramachandran | G. Ramanathan | Krishna Pictures |\n| 1961 | Batasari | Telugu | Madhavi | A. Nageswara Rao | Master Venu | Bharani Pictures |\n| 1961 | Kaanal Neer | Tamil | Madhavi | A. Nageswara Rao | Master Venu | Bharani Pictures |\n| 1962 | Annai | Tamil | Savithri | S. V. Ranga Rao | R. Sudharsanam | AVM Productions |\n| 1963 | Kalai Arasi | Tamil | Vani & Valli | M. G. Ramachandran | K. V. Mahadevan | Sarodi Brothers |\n| 1963 | Anuragam | | | | | |\n| 1963 | Arivaali | Tamil | Manorama | Sivaji Ganesan | S. V. Venkatraman | A. T. K. Productions |\n| 1963 | Penchina Prema | Telugu | | | | |\n| 1963 | Kaanchi Thalaivan | Tamil | | M. G. Ramachandran | K. V. Mahadevan | Mekala Pictures |\n| 1964 | Bobbili Yudham | Telugu | Mallamma | N. T. Rama Rao | S. Rajeswara Rao | |\n| 1964 | Vivaha Bandham | Telugu | Bharati | | | |\n| 1965 | Antastulu | Telugu | | A. Nageswara Rao | K. V. Mahadevan | |\n| 1965 | Todu Needa | Telugu | | | | |\n| 1966 | Palnati Yudham | Telugu | Nagamma | N. T. Rama Rao | S. Rajeswara Rao | Sri Anuroopa Films |\n| 1967 | Grihalakshmi | Telugu | Laxmi | | | |\n| 1967 | Pattatthu Rani | Tamil | | Gemini Ganesan | T. K. Ramamoorthy | |\n| 1968 | Poovum Pottum | Tamil | Padmavathi | S. V. Ranga Rao | R. Govardhanam | Vasu Films |\n| 1971 | Mattilo Manikyam | Telugu | Lalitha | | | |\n| 1972 | Anta Mana Manchike | Telugu | Savitri | | | |\n| 1973 | Kattila Thottila | Tamil | | Gemini Ganesan | | Arul Films |\n| 1973 | Vichitra Vivaham | | | | | |\n| 1974 | Ammayi Pelli | Telugu | | | | |\n| 1974 | Tatamma Kala | Telugu | | | | |\n| 1975 | Ippadiyum Oru Penn | Tamil | | | | |\n| 1975 | Pathu Maadha Bandham | Tamil | S. A. Asokan | Krishnan-Panju | | |\n| 1977 | Manavadi Kosam | Telugu | | | | |\n| 1984 | Mangamma Gari Manavadu | Telugu | Mangamma | | | |\n| 1986 | Muddula Manavaraalu | Telugu | | | | |\n| 1986 | Kannukku Mai Ezhuthu | Tamil | | Sujatha, Sarath Babu | Ilaiyaraaja | |\n| 1988 | Attagaru Swagatham | Telugu | | | | |\n| 1990 | Bamma Maata Bangaru Baata | Telugu | Bamma | | | |\n| 1992 | Chamanthi | Telugu | Bhuvaneswari | | | |\n| 1992 | Peddarikam | Telugu | Adusumilli Basavapunnamma | | | |\n| 1992 | Samrat Ashok | Telugu | | | | |\n| 1992 | Chembaruthi | Tamil | | Prashanth, Roja | Ilaiyaraaja | Motherland Movies |\n| 1998 | Pelli Kanuka | Telugu | Savithramma | | | |\n\n\nQuestion\nwhat was the name of the last movie the actress appeared in?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"Pelli Kanuka"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-9576",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-3111.csv` in your code execution environment.\n\nTable Title: Franz Liszt's treatments of the works of other composers\nTable Content:\n| Original work | Liszt work | Forces | Date | S no. | Notes |\n| --- | --- | --- | --- | --- | --- |\n| Opera Lucrezia Borgia (1833) | Reminiscences de Lucrezia Borgia | piano | 1840 | S.400 | The sections are: (1) Trio du second acte and (2) Fantaisie sur des motifs favoris de l'opéra: Chanson à boire (Orgie), Duo and Finale |\n| Opera Lucia di Lammermoor (1835) | Reminiscences de Lucia di Lammermoor | piano | 1835‑36 | S.397 | Based on the Sextet |\n| Opera Lucia di Lammermoor (1835) | Marche et cavatine de Lucie de Lammermoor | piano | 1835‑36 | S.398 | Intended as part of the Reminiscences (S.397) but separated by the publisher |\n| Lucia di Lammermoor | Valse à capriccio sur deux motifs de Lucia et Parisina | piano | 1842 | S.401 | Combines motifs from Lucia di Lammermoor and Parisina. The Valse à capriccio, S.401 was revised as No. 3 of 3 Caprices-Valses, S.214 (1850–52). |\n| Opera Parisina (1833) | Valse à capriccio sur deux motifs de Lucia et Parisina | piano | 1842 | S.401 | Combines motifs from Lucia di Lammermoor and Parisina. The Valse à capriccio, S.401 was revised as No. 3 of 3 Caprices-Valses, S.214 (1850–52). |\n| Nuits d'été à Pausilippe, 12 ariettas and nocturnes (voices and chamber ensemble; 1836):\\n\\nNo. 1, \"Barcajuolo\"\\nNo. 9, \"L'Alito di Bice\"\\nNo. 4, \"La Torre di Biasone\" | Nuits d'été à Pausilippe | piano | 1838 | S.399 | |\n| Opera La favorite (1840):\\n\\nDuettino (Act IV) | Spirito gentil | piano | 1847 | S.400a | Possibly same as S.742. |\n| Opera Dom Sébastien (1843):\\n\\nFuneral march | Marche funèbre de Dom Sébastien | piano | 1844 | S.402 | |\n| Act IV | Paraphrase | piano | ? | S.744 | Lost; this was based on Theodor Kullak's transcription (paraphrase), Op. 31, which was itself dedicated to Liszt. |\n| \"Duettino\" | Transcription | piano | ? | S.742 | Lost. Possibly same as S.400a. |\n\n\nQuestion\nwas opera dom sebastien of opera la favorite published in 1844?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"yes"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-3111",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-12691.csv` in your code execution environment.\n\nTable Title: Novi Sad raid\nTable Content:\n| Place | Total | Men | Women | Children | Elderly | Serbs | Jews | Romani | Rusyns | Hungarians | Russians |\n| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |\n| Bečej | 215 | 111 | 72 | 13 | 19 | 102 | 110 | - | - | - | - |\n| Vilovo | 64 | 44 | 6 | 8 | 6 | 64 | - | - | - | - | - |\n| Gardinovci | 37 | 32 | 3 | - | 2 | 37 | - | - | - | - | - |\n| Gospođinci | 85 | 47 | 19 | 15 | 4 | 73 | 10 | - | 2 | - | - |\n| Đurđevo | 223 | 107 | 60 | 41 | 15 | 173 | 22 | - | 27 | - | - |\n| Žabalj | 666 | 355 | 141 | 101 | 69 | 614 | 28 | 23 | - | 1 | - |\n| Lok | 47 | 46 | - | - | 1 | 46 | - | - | - | 1 | - |\n| Mošorin | 205 | 94 | 41 | 44 | 26 | 170 | - | 34 | - | 1 | - |\n| Novi Sad | 1,246 | 489 | 415 | 165 | 177 | 375 | 809 | - | 2 | 18 | 15 |\n| Srbobran | 3 | 3 | - | - | - | 2 | 2 | - | - | - | - |\n| Temerin | 48 | 14 | 15 | 7 | 12 | 6 | 42 | - | - | - | - |\n| Titel | 51 | 45 | - | 1 | 5 | 49 | 1 | - | - | - | - |\n| Čurug | 893 | 554 | 153 | 82 | 104 | 842 | 44 | 7 | - | - | - |\n| Šajkaš | 26 | 24 | 2 | - | - | 25 | 1 | - | - | - | - |\n| All places | 3,809 | 1,965 | 927 | 477 | 440 | 2,578 | 1,068 | 64 | 31 | 21 | 15 |\n\n\nQuestion\nwhich place has the most women?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"Novi Sad"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-12691",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-4421.csv` in your code execution environment.\n\nTable Title: Parnelli Jones\nTable Content:\n| Year | Car | Start | Qual | Rank | Finish | Laps | Led | Retired |\n| --- | --- | --- | --- | --- | --- | --- | --- | --- |\n| 1961 | 98 | 5 | 146.080 | 7 | 12 | 192 | 27 | Flagged |\n| 1962 | 98 | 1 | 150.370 | 1 | 7 | 200 | 120 | Running |\n| 1963 | 98 | 1 | 151.153 | 1 | 1 | 200 | 167 | Running |\n| 1964 | 98 | 4 | 155.099 | 4 | 23 | 55 | 7 | Pit fire |\n| 1965 | 98 | 5 | 158.625 | 5 | 2 | 200 | 0 | Running |\n| 1966 | 98 | 4 | 162.484 | 4 | 14 | 87 | 0 | Wheel Bearing |\n| 1967 | 40 | 6 | 166.075 | 6 | 6 | 196 | 171 | Bearing |\n| Totals | Totals | Totals | Totals | Totals | Totals | 1130 | 492 | |\n\n\nQuestion\nin how many indy 500 races, has jones been flagged?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"1"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-4421",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-13978.csv` in your code execution environment.\n\nTable Title: Svetoslav Todorov\nTable Content:\n| # | Date | Venue | Opponent | Score | Result | Competition |\n| --- | --- | --- | --- | --- | --- | --- |\n| 1. | 29 March 2000 | Vasil Levski National Stadium, Sofia | Belarus | 1–0 | 4–1 | Friendly |\n| 2. | 7 October 2000 | Georgi Asparuhov Stadium, Sofia | Malta | 3–0 | 3–0 | 2002 World Cup qualifier |\n| 3. | 27 March 2003 | Stadion Mladost, Kruševac | Serbia and Montenegro | 2–1 | 2–1 | Friendly |\n| 4. | 7 June 2003 | Vasil Levski National Stadium, Sofia | Belgium | 2–2 | 2–2 | Euro 2004 qualifier |\n| 5. | 12 November 2005 | Vasil Levski National Stadium, Sofia | Georgia | 5–0 | 6–2 | Friendly |\n| 6. | 12 November 2005 | Vasil Levski National Stadium, Sofia | Georgia | 6–2 | 6–2 | Friendly |\n| 7. | 9 May 2006 | Nagai Stadium, Osaka | Japan | 1–0 | 2–1 | 2006 Kirin Cup |\n\n\nQuestion\nwho was the opponent after georgia?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"Japan"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-13978",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-5925.csv` in your code execution environment.\n\nTable Title: IWGP Heavyweight Championship\nTable Content:\n| Rank | Wrestler | No. of reigns | Combined\\ndefenses | Combined\\ndays |\n| --- | --- | --- | --- | --- |\n| 1 | Keiji Mutoh | 4 | 19 | 1,238 |\n| 2 | Hiroshi Tanahashi | 6 | 27 | 1,237 |\n| 3 | Shinya Hashimoto | 3 | 20 | 1,052 |\n| 4 | Tatsumi Fujinami | 6 | 13 | 785 |\n| 5 | Kensuke Sasaki | 5 | 9 | 647 |\n| 6 | Yuji Nagata | 2 | 12 | 570 |\n| 7 | Kazuchika Okada† | 2 | 10 | 556+ |\n| 8 | Kazuyuki Fujita | 3 | 3 | 478 |\n| 9 | Big Van Vader | 3 | 4 | 451 |\n| 10 | Shinsuke Nakamura | 3 | 9 | 390 |\n| 11 | Riki Chōshū | 3 | 5 | 383 |\n| 12 | Antonio Inoki | 1 | 4 | 325 |\n| 13 | Brock Lesnar | 1 | 3 | 280 |\n| 14 | Hiroyoshi Tenzan | 4 | 2 | 197 |\n| 15 | Yoshihiro Takayama | 1 | 3 | 185 |\n| 16 | Satoshi Kojima | 2 | 2 | 168 |\n| 17 | Togi Makabe | 1 | 3 | 161 |\n| 18 | Scott Norton | 2 | 4 | 126 |\n| 19 | Nobuhiko Takada | 1 | 1 | 116 |\n| 20 | Bob Sapp | 1 | 1 | 66 |\n| 21 | Salman Hashimikov | 1 | 0 | 48 |\n| 21 | Tadao Yasuda | 1 | 1 | 48 |\n| 23 | Manabu Nakanishi | 1 | 0 | 45 |\n| 24 | Masahiro Chono | 1 | 0 | 44 |\n| 25 | Genichiro Tenryu | 1 | 0 | 25 |\n\n\nQuestion\neach of these wrestlers came from a reign of 3\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"Shinya Hashimoto",
"Kazuyuki Fujita",
"Big Van Vader",
"Shinsuke Nakamura",
"Riki Chōshū"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-5925",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-13836.csv` in your code execution environment.\n\nTable Title: 1998 Marlboro Grand Prix of Miami\nTable Content:\n| Pos | No | Driver | Team | Laps | Time/Retired | Grid | Points |\n| --- | --- | --- | --- | --- | --- | --- | --- |\n| 1 | 6 | Michael Andretti | Newman-Haas Racing | 150 | 1:33:39.268 | 8 | 20+1 |\n| 2 | 99 | Greg Moore | Forsythe Racing | 150 | +0.075 | 1 | 16+1 |\n| 3 | 1 | Alex Zanardi | Chip Ganassi Racing | 150 | +0.918 | 6 | 14 |\n| 4 | 11 | Christian Fittipaldi | Newman-Haas Racing | 150 | +3.359 | 4 | 12 |\n| 5 | 20 | Scott Pruett | Patrick Racing | 150 | +5.257 | 12 | 10 |\n| 6 | 40 | Adrián Fernández | Patrick Racing | 150 | +5.926 | 5 | 8 |\n| 7 | 5 | Gil de Ferran | Walker Racing | 150 | +6.810 | 11 | 6 |\n| 8 | 8 | Bryan Herta | Team Rahal | 150 | +9.866 | 16 | 5 |\n| 9 | 27 | Dario Franchitti | Team KOOL Green | 150 | +10.275 | 13 | 4 |\n| 10 | 17 | Maurício Gugelmin | PacWest Racing Group | 150 | +10.855 | 10 | 3 |\n| 11 | 33 | Patrick Carpentier | Forsythe Racing | 150 | +11.382 | 9 | 2 |\n| 12 | 18 | Mark Blundell | PacWest Racing Group | 150 | +29.029 | 7 | 1 |\n| 13 | 10 | Richie Hearn | Della Penna Motorsports | 149 | +1 Lap | 15 | |\n| 14 | 9 | JJ Lehto | Hogan Racing | 149 | +1 Lap | 29 | |\n| 15 | 15 | Roberto Moreno | Project Indy | 149 | +1 Lap | 23 | |\n| 16 | 12 | Jimmy Vasser | Chip Ganassi Racing | 148 | +2 Laps | 3 | |\n| 17 | 3 | André Ribeiro | Marlboro Team Penske | 148 | +2 Laps | 2 | |\n| 18 | 36 | Alex Barron | All American Racing | 148 | +2 Laps | 27 | |\n| 19 | 7 | Bobby Rahal | Team Rahal | 147 | +3 Laps | 17 | |\n| 20 | 98 | P. J. Jones | All American Racing | 146 | +4 Laps | 26 | |\n| 21 | 77 | Arnd Meier | Davis Racing | 136 | +14 Laps | 28 | |\n| 22 | 2 | Al Unser, Jr. | Marlboro Team Penske | 122 | Transmission | 14 | |\n| 23 | 24 | Hiro Matsushita | Arciero-Wells Racing | 121 | Contact | 24 | |\n| 24 | 16 | Hélio Castro-Neves | Bettenhausen Racing | 97 | Contact | 22 | |\n| 25 | 34 | Dennis Vitolo | Payton/Coyne Racing | 90 | Withdrew | 25 | |\n| 26 | 25 | Max Papis | Arciero-Wells Racing | 83 | Wheel Bearing | 20 | |\n| 27 | 26 | Paul Tracy | Team KOOL Green | 79 | Suspension | 21 | |\n| 28 | 19 | Michel Jourdain Jr. | Payton/Coyne Racing | 78 | Handling | 19 | |\n| 29 | 21 | Tony Kanaan | Tasman Motorsports Group | 32 | Contact | 18 | |\n\n\nQuestion\nhow many teams competed in the 1998 marlboro grand prix of miami?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"29"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-13836",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-2493.csv` in your code execution environment.\n\nTable Title: Shooting at the 1988 Summer Olympics – Women's 50 metre rifle three positions\nTable Content:\n| Rank | Athlete | Qual | Final | Total | Notes |\n| --- | --- | --- | --- | --- | --- |\n| | Silvia Sperber (FRG) | 590 | 95.6 | 685.6 | OR |\n| | Vesela Letcheva (BUL) | 583 | 100.2 | 683.2 | |\n| | Valentina Cherkasova (URS) | 586 | 95.4 | 681.4 | |\n| 4 | Katja Klepp (GDR) | 584 | 96.5 | 680.5 | |\n| 5 | Sharon Bowes (CAN) | 584 | 96.5 | 680.5 | |\n| 6 | Anna Maloukhina (URS) | 585 | 93.4 | 678.4 | |\n| 7 | Launi Meili (USA) | 582 | 94.5 | 676.5 | |\n| 8 | Anita Karlsson (SWE) | 583 | 93.4 | 676.4 | |\n\n\nQuestion\nwhich runners ran above 95.4?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"Silvia Sperber (FRG)",
"Vesela Letcheva (BUL)",
"Katja Klepp (GDR)",
"Sharon Bowes (CAN)"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-2493",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-3586.csv` in your code execution environment.\n\nTable Title: List of Royal Military College of Canada people\nTable Content:\n| Year | Number | Name | Year | Number | Name | Year | Number | Name |\n| --- | --- | --- | --- | --- | --- | --- | --- | --- |\n| 1884–1885 | 7 | Lukin Homphrey Irving (first) | 1886–1887 | 18 | Duncan MacPherson | 1888 | 4 | William Mahlon Davis |\n| 1889–1890 | 6 | Septimus Julius Augustus Denison | 1891 | 10 | Victor Brereton Rivers | 1892 | 86 | Reuben Wells Leonard |\n| 1893–1894 | 37 | E.H. Drury | 1895–1896 | 15 | Francis Joseph Dixon | 1897 | 48 | A.K. Kirkpatrick |\n| 1898 | 57 | H.S. Greenwood | 1899 | 14 | John Bray Cochrane | 1900 | 41 | Robert Cartwright |\n| 1901 | 154 | F.M. Gaudet | 1902 | 47 | Ernest Frederick Wurtele | 1903 | 21 | A.E. Doucet |\n| 1904 | 82 | Wallace Bruce Matthews Carruthers | 1905 | 188 | W.A.H. Kerr | 1906 | 186 | V.A.S. Williams |\n| 1907 | 139 | C.R.F. Coutlee | 1908 | 232 | John Houlison | 1909 | 91 | J.D. Gibson |\n| 1910 | 63 | George Hooper | 1911 | 255 | H.A. Panet | 1912 | 246 | Major-General Sir Henry Edward Burstall |\n| 1913 | 268 | Henry Robert Visart de Bury et de Bocarmé | 1914; 1919 | 299 | Col. Harry J. Lamb DSO, VD | 1920 | 293 | C.J. Armstrong |\n| 1920–1922 | 392 | W.B. Kingsmill | 1923 | 377 | A.C. Caldwell | 1924 | 140 | G.S. Cartwright |\n| 1925 | 499 | Edouard de B. Panet | 1926 | 631 | A.B. Gillies | 1927 | 623 | S.B. Coristine |\n| 1928 | 555 | R.R. Carr-Harris | 1929 | 667 | E.G. Hanson | 1929–1930 | SUO | G.D. de S. Wotherspoon |\n| 1930–1931 | 1119 | J.H. Price | 1932 | 472 | A.R. Chipman | 1933–1934 | 805 | Colin W. G. Gibson |\n| 1935 | 727 | D.A. White | 1936–1937 | 877 | G.L. Magann | 1938–1939 | 1003 | A.M. Mitchell |\n| 1940–1941 | 803 | J.V. Young | 1942–1943 | 1141 | W.H. O'Reilly | 1944 | 698 | Everett Bristol |\n| 1945 | 982 | D.W. MacKeen | 1946 | 1841 | D.G. Cunningham | 1947 | 1230 | S.H. Dobell |\n| 1948 | 1855 | Ian S. Johnston | 1949 | 1625 | J.D. Watt | 1950 | 1542 | E.W. Crowe |\n| 1951 | 1860 | Nicol Kingsmill | 1952 | 1828 | Ted G.E. Beament | 1953 | 1620 | R.R. Labatt |\n| 1954 | 1766 | Ken H. Tremain | 1955 | 1474 | de L.H.M Panet | 1956 | 2034 | Paul Y. Davoud |\n| 1957 | 1954 | W.P. Carr | 1960 | 1379 | H.A. Mackenzie | 1961 | 2157 | J.H.R. Gagnon |\n| 1962 | 2183 | James E. Pepall | 1963 | 2336 | J.H. Moore | 1964 | 2351 | Guy Savard |\n| 1965 | 2749 | James B. Cronyn | 1966 | 2601 | J. Fergus Maclaren | 1967 | 2791 | Jean P.W. Ostiguy |\n| 1968–1969 | RCNC90 | John F. Frank | 1975–1976 | 3661 | Terry Yates | 1976–1977 | 5533 | Glenn Allen |\n| 1977–1978 | 3172 | Marshall Soule | 1980–1981 | 3251 | Jim Tremain | 1981–1982 | 2897 | Herb Pitts |\n| 1986–1987 | 5604 | Ken Smee | 1987–1988 | 3010 | Peter McLoughlin | 1992–1993 | H3356 | Robin Cumine |\n| 1993–1994 | 5244 | Tony Downs | 1994–1995 | H7543 | Senator Joseph A. Day | 1995–1996 | 5739 | Andre Costin |\n| 1996–1997 | 3550 | Murray Johnston | 1997–1998 | 8813 | John D. Gibson | 1998–1999 | G0055 | Valerie Keyes (first female) |\n| 1999–2000 | 8833 | John Leggat | 2000–2001 | 5758 | Michael Morres | 2001–2002 | 16461 | Ian MacKinnon |\n| 2002–2003 | 6777 | Michel Charron | 2003–2004 | 7776 | Chris Lythgo | 2004–2005 | 7943 | J. William K. Lye |\n| 2005–2006 | 10080 | Robert Booth | 2007–2008 | 6776 | Tim Sparling | 2008–2009 | 15988 | Jeff Kearns |\n| 2010 | 16412 | Gord Clarke | 2011 | 19307 | David Benoit | 2012 | 9889 | Robert Benn |\n| 2013 | M0058 | Marc Drolet (first UTPNCM) | | | | | | |\n\n\nQuestion\nhow many club presidents held their position in the 19th century?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"4"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-3586",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-4899.csv` in your code execution environment.\n\nTable Title: Neha Sharma\nTable Content:\n| Year | Film | Role | Language | Notes |\n| --- | --- | --- | --- | --- |\n| 2007 | Chirutha | Sanjna | Telugu | Debut Telugu Film |\n| 2009 | Kurradu | Hema | Telugu | |\n| 2010 | Crook: It's Good To Be Bad | Suhani | Hindi | Debut Hindi Film |\n| 2012 | Teri Meri Kahaani | Meera | Hindi | Special Appearance |\n| 2012 | Kyaa Super Kool Hain Hum | Simran | Hindi | |\n| 2013 | Jayantabhai Ki Luv Story | Simran | Hindi | |\n| 2013 | Yamla Pagla Deewana 2 | Suman | Hindi | |\n| 2014 | Youngistaan | Anwita Chauhan | Hindi | |\n\n\nQuestion\nwhat film game before ter meri kahaani?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"Crook: It's Good To Be Bad"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-4899",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-13437.csv` in your code execution environment.\n\nTable Title: Pillow Pal\nTable Content:\n| Name | Animal Type | Introduced | Reintroduced | Retired | Beanie Baby Resembled |\n| --- | --- | --- | --- | --- | --- |\n| Antlers | Moose | | 1999 (neon) | 1999 | Chocolate |\n| Baba | Lamb | | 1999 (neon) | 1999 | Fleece |\n| Bruiser | Bulldog | 1997 | | | Wrinkles |\n| Carrots | Bunny | | 1999 (neon) | 1999 | Ears, but pink instead of brown |\n| Chewy | | | 1999 (neon) | 1999 | |\n| Clover | Bunny | 1998 | | | Ears, but white instead of brown |\n| Foxy | Fox | | | | Sly |\n| Glide | Porpoise | | | | Echo the Dolphin |\n| Huggy | Bear | 1995 | 1999 (neon) | 1999 | Cubbie, but yellow instead of brown |\n| Meow | Cat | 1997 | 1999 (neon) | 1999 | Snip |\n| Moo | Cow | 1995 | | | Daisy, but with white body instead of black |\n| Oink | Pig | 1995 | | | Squealer |\n| Paddles | Platypus | | 1999 (neon) | 1999 | Patti |\n| Purr | Tiger | | | | Stripes, but yellow-orange instead of orange-black |\n| Red | Bull | | | | Snort |\n| Ribbit (All-Green) | Frog | 1995 | 1999 (neon) | 1999 | Legs |\n| Ribbit (Green-Yellow) | Frog | | | | Smoochy |\n| Sherbert | Tie-Dye Bear | 1998 | 1999 (neon) | 1999 | Sammy |\n| Snap (All-Yellow) | Turtle | 1995 | | | Speedy, but yellow instead of green |\n| Snap (Yellow-Green) | Turtle | | 1999 (neon) | | Peekaboo |\n| Snuggy | Bear | 1995 | | | Cubbie, but pink instead of brown |\n| Sparkler | Bear | | | 1999 | Spangle |\n| Speckles | Leopard | | | | Freckles |\n| Spotty | Dalmatian | | | | Dotty |\n| Squirt | Elephant | | 1999 (neon) | 1999 | Peanut |\n| Swinger | Monkey | | 1999 (neon) | 1999 | Bongo |\n| Tide | Whale | | | | Splash |\n| Tubby | Hippopotamus | | | | Happy |\n| Woof | Dog | 1995 | 1999 (neon) | 1999 | Bones |\n| Zulu | Zebra | | | | Ziggy, but pink-green instead of black-white |\n\n\nQuestion\nwhat is the name of the pillow pal listed after clover?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"Foxy"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-13437",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-5586.csv` in your code execution environment.\n\nTable Title: Wikimania\nTable Content:\n| Conference | Date | Place | attendance | Archive of presentations |\n| --- | --- | --- | --- | --- |\n| Wikimania 2005 | August 5–7 | Frankfurt, Germany | 380 | slides, video |\n| Wikimania 2006 | August 4–6 | Cambridge, United States | 400 | slides and papers, video |\n| Wikimania 2007 | August 3–5 | Taipei, ROC (Taiwan) | 440 | Commons gallery |\n| Wikimania 2008 | July 17–19 | Alexandria, Egypt | 650 | abstracts, slides,video |\n| Wikimania 2009 | August 26–28 | Buenos Aires, Argentina | 559 | slides, video |\n| Wikimania 2010 | July 9–11 | Gdańsk, Poland | about 500 | slides |\n| Wikimania 2011 | August 4–7 | Haifa, Israel | 720 | presentations, video |\n| Wikimania 2012 | July 12–15 | Washington, D.C., United States | 1,400 | presentations, videos |\n| Wikimania 2013 | August 7–11 | Hong Kong | 700 | presentations, videos |\n| Wikimania 2014 | August 6–10 | London, United Kingdom | N/A | |\n\n\nQuestion\nwhere was the conference held the year after buenos aires, argentina?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"Gdańsk, Poland"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-5586",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-1127.csv` in your code execution environment.\n\nTable Title: Smile! (Vitas album)\nTable Content:\n| Track number | Russian title | English title | Music | Lyrics |\n| --- | --- | --- | --- | --- |\n| 01 | Улыбнись | Smile! | Vitas | D. Plachkovsky |\n| 02 | Блаженный Гуру | Blessed Guru | Vitas | Vitas |\n| 03 | Восковые фигуры | Wax Figures | Vitas | D. Plachkovsky, Vitas, S. Pudovkin |\n| 04 | Плачет чужая тоска | Someone's Melancholy Is Crying | Vitas | D. Plachovsky |\n| 05 | Аве Мария | Ave Maria | Franz Schubert | Traditional prayer |\n| 06 | Холодный мир | The Cold World | V. Molchanov | D. Plachkovsky |\n| 07 | Остров затонувших кораблей | The Island Of Sunken Ships | O. Gazmanov | E. Nebylova |\n| 08 | Фантастические сны | Fantastic Dreams | Vitas | Vitas, D. Plachkovsky |\n| 09 | Good-bye | Good-bye | Vitas | Vitas |\n| 10 | Счастье | Happiness | Vitas | Vitas |\n| 11 | Отцвели хризантемы | Chrysanthemums Have Faded Out | N. Harito | V. Shumsky |\n| 12 | Дождь в Тбилиси | Rain in Tbilisi | Vitas | Ratner |\n| 13 | Слепой художник | Blind Painter | Vitas | D. Plachkovsky, Vitas, S. Pudovkin |\n| 14 | Милая музыка | Lovely Music | Vitas | Vitas |\n| 15 | Белоруссия | Byelorussia | Alexandra Pakhmutova | N. Dobronravov |\n| 16 | До свидания | See You Later (Do Svidaniya) | Vitas | Vitas |\n\n\nQuestion\nwhich lyricist has the largest amount of tracks?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"Vitas"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-1127",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-11317.csv` in your code execution environment.\n\nTable Title: IFK Visby\nTable Content:\n| Season | Level | Division | Section | Position | Movements |\n| --- | --- | --- | --- | --- | --- |\n| 1999 | Tier 5 | Division 4 | Gotland | 6th | Vårserier (Spring Series) |\n| | Tier 5 | Division 4 | Gotland | 5th | Höstserier (Autumn Series) |\n| 2000 | Tier 5 | Division 4 | Gotland | 2nd | Vårserier (Spring Series) |\n| | Tier 5 | Division 4 | Uppland/Gotland Höst | 10th | Höstserier (Autumn Series) |\n| 2001 | Tier 5 | Division 4 | Gotland | 4th | Vårserier (Spring Series) |\n| | Tier 5 | Division 4 | Gotland | 4th | Slutspel A (Playoff A) |\n| 2002 | Tier 5 | Division 4 | Gotland | 3rd | Promotion Playoffs |\n| 2003 | Tier 5 | Division 4 | Gotland | 6th | |\n| 2004 | Tier 5 | Division 4 | Gotland | 4th | |\n| 2005 | Tier 5 | Division 4 | Gotland | 1st | |\n| 2006* | Tier 5 | Division 3 | Östra Svealand | 12th | Relegated |\n| 2007 | Tier 6 | Division 4 | Gotland | 2nd | |\n| 2008 | Tier 6 | Division 4 | Gotland | 3rd | |\n| 2009 | Tier 6 | Division 4 | Gotland | 7th | |\n| 2010 | Tier 6 | Division 4 | Gotland | 2nd | |\n| 2011 | Tier 6 | Division 4 | Gotland | 4th | |\n\n\nQuestion\nwhat is the total number of times that the team finished in 3rd position or higher?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"6"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-11317",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-2000.csv` in your code execution environment.\n\nTable Title: Electoral system of Australia\nTable Content:\n| Seat | State | Majority | Member | Party |\n| --- | --- | --- | --- | --- |\n| HINKLER | Qld | 00.0 | Paul Neville | NPA |\n| SOLOMON | NT | 00.1 | Dave Tollner | Lib |\n| ADELAIDE | SA | 00.2 | Hon Trish Worth | Lib |\n| CANNING | WA | 00.4 | Don Randall | Lib |\n| DOBELL | NSW | 00.4 | Ken Ticehurst | Lib |\n| PARRAMATTA | NSW | 01.1 | Ross Cameron | Lib |\n| McEWEN | Vic | 01.2 | Fran Bailey | Lib |\n| PATERSON | NSW | 01.4 | Bob Baldwin | Lib |\n| HERBERT | Qld | 01.6 | Peter Lindsay | Lib |\n| RICHMOND | NSW | 01.6 | Hon Larry Anthony | NPA |\n| DEAKIN | Vic | 01.7 | Philip Barresi | Lib |\n| EDEN-MONARO | NSW | 01.7 | Gary Nairn | Lib |\n| HINDMARSH | SA | 01.9 | Hon Christine Gallus | Lib |\n\n\nQuestion\nwhat was the total majority that the dobell seat had?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"00.4"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-2000",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-8986.csv` in your code execution environment.\n\nTable Title: Connors–McEnroe rivalry\nTable Content:\n| Year | Tournament | Surface | Round | Winner | Score |\n| --- | --- | --- | --- | --- | --- |\n| 1991 | Basel | Hard | S | John McEnroe | 6–1, 6–3 |\n| 1989 | Toulouse | Hard | F | Jimmy Connors | 6–3, 6–3 |\n| 1987 | Montreal / Toronto | Hard | Q | Jimmy Connors | 6–3, 3–6, 6–3 |\n| 1986 | San Francisco | Carpet | F | John McEnroe | 7–6, 6–3 |\n| 1985 | Montreal / Toronto | Hard | S | John McEnroe | 6–2, 6–3 |\n| 1985 | Chicago | Carpet | F | John McEnroe | W/O |\n| 1984 | US Open | Hard | S | John McEnroe | 6–4, 4–6, 7–5, 4–6, 6–3 |\n| 1984 | Montreal / Toronto | Hard | S | John McEnroe | 2–6, 6–2, 6–3 |\n| 1984 | Wimbledon | Grass | F | John McEnroe | 6–1, 6–1, 6–2 |\n| 1984 | London, Queen's Club | Grass | S | John McEnroe | 6–2, 6–2 |\n| 1984 | Roland Garros | Clay | S | John McEnroe | 7–5, 6–1, 6–2 |\n| 1984 | Dallas WCT | Carpet | F | John McEnroe | 6–1, 6–2, 6–3 |\n| 1983 | Wembley | Carpet | F | John McEnroe | 7–5, 6–1, 6–4 |\n| 1983 | Cincinnati | Hard | S | John McEnroe | 6–7, 6–1, 6–4 |\n| 1983 | London, Queen's Club | Grass | F | Jimmy Connors | 6–3, 6–3 |\n| 1982 | San Francisco | Carpet | F | John McEnroe | 6–1, 6–3 |\n| 1982 | Wimbledon | Grass | F | Jimmy Connors | 3–6, 6–3, 6–7, 7–6, 6–4 |\n| 1982 | London, Queen's Club | Grass | F | Jimmy Connors | 7–5, 6–3 |\n| 1982 | Philadelphia | Carpet | F | John McEnroe | 6–3, 6–3, 6–1 |\n| 1981 | New York City | Carpet | RR | John McEnroe | 6–2, 7–5 |\n| 1981 | Wembley | Carpet | F | Jimmy Connors | 3–6, 2–6, 6–3, 6–4, 6–2 |\n| 1980 | US Open | Hard | S | John McEnroe | 6–4, 5–7, 0–6, 6–3, 7–6 |\n| 1980 | Wimbledon | Grass | S | John McEnroe | 6–3, 3–6, 6–3, 6–4 |\n| 1980 | Dallas WCT | Carpet | F | Jimmy Connors | 2–6, 7–6, 6–1, 6–2 |\n| 1980 | Memphis | Carpet | F | John McEnroe | 7–6, 7–6 |\n| 1980 | Philadelphia | Carpet | F | Jimmy Connors | 6–3, 2–6, 6–3, 3–6, 6–4 |\n| 1979 | US Open | Hard | S | John McEnroe | 6–3, 6–3, 7–5 |\n| 1979 | Dallas WCT | Carpet | S | John McEnroe | 6–1, 6–4, 6–4 |\n| 1979 | Las Vegas | Hard | S | Jimmy Connors | 7–5, 6–4 |\n| 1979 | Pepsi Grand Slam | Clay | S | Jimmy Connors | 6–3, 6–4 |\n| 1978 | Masters | Carpet | RR | John McEnroe | 7–5, 3–0 RET |\n| 1978 | US Open | Hard | S | Jimmy Connors | 6–2, 6–2, 7–5 |\n| 1978 | Indianapolis | Clay | Q | Jimmy Connors | 3–6, 6–1, 6–1 |\n| 1977 | Boston | Clay | R32 | Jimmy Connors | 5–7, 6–2, 7–5 |\n| 1977 | Wimbledon | Grass | S | Jimmy Connors | 6–3, 6–3, 4–6, 6–4 |\n\n\nQuestion\nwhat is the total number of carpet surfaces used?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"13"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-8986",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-3151.csv` in your code execution environment.\n\nTable Title: 2009 Utah Utes football team\nTable Content:\n| Date | Time | Opponent# | Rank# | Site | TV | Result | Attendance |\n| --- | --- | --- | --- | --- | --- | --- | --- |\n| September 3 | 7:00 PM | Utah State* | #19 | Rice–Eccles Stadium • Salt Lake City, UT (Battle of the Brothers) | The mtn. | W 35–17 | 45,333 |\n| September 12 | 8:30 PM | at San Jose State* | #17 | Spartan Stadium • San José, CA | ESPNU | W 24–14 | 23,684 |\n| September 19 | 1:30 PM | at Oregon* | #18 | Autzen Stadium • Eugene, OR | ESPN | L 24–31 | 58,017 |\n| September 26 | 5:30 PM | Louisville* | | Rice–Eccles Stadium • Salt Lake City, UT | CBSCS | W 30–14 | 45,588 |\n| October 10 | 4:00 PM | at Colorado State | | Hughes Stadium • Fort Collins, CO | The mtn. | W 24–17 | 30,499 |\n| October 17 | 8:00 PM | at UNLV | #24 | Sam Boyd Stadium • Las Vegas, NV | The mtn. | W 35–15 | 26,315 |\n| October 24 | 2:00 PM | Air Force | #19 | Rice–Eccles Stadium • Salt Lake City, UT | Versus | W 23–16 OT | 45,129 |\n| October 31 | 6:00 PM | Wyoming | #19 | Rice–Eccles Stadium • Salt Lake City, UT | The mtn. | W 22–10 | 44,837 |\n| November 7 | 4:00 PM | New Mexico | #17 | Rice–Eccles Stadium • Salt Lake City, UT (The Bowl) | The mtn. | W 45–14 | 45,051 |\n| November 14 | 5:30 PM | at #4 TCU | #16 | Amon G. Carter Stadium • Fort Worth, TX | CBSCS | L 28–55 | 50,307 |\n| November 21 | 2:00 PM | San Diego State | #23 | Rice–Eccles Stadium • Salt Lake City, UT | Versus | W 38–7 | 44,991 |\n| November 28 | 3:00 PM | at #18 BYU | #22 | LaVell Edwards Stadium • Provo, UT (The Holy War) | CBSCS/The mtn. | L 23–26 OT | 64,301 |\n| December 23 | 6:00 PM | vs. California* | #23 | Qualcomm Stadium • San Diego, CA (Poinsettia Bowl) | ESPN | W 37–27 | 32,665 |\n\n\nQuestion\nagainst what opponent did utah score the largest amount of points?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"New Mexico"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-3151",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-4857.csv` in your code execution environment.\n\nTable Title: 2008 Malaysian motorcycle Grand Prix\nTable Content:\n| Pos | No | Rider | Manufacturer | Laps | Time | Grid | Points |\n| --- | --- | --- | --- | --- | --- | --- | --- |\n| 1 | 46 | Valentino Rossi | Yamaha | 21 | 43:06.007 | 2 | 25 |\n| 2 | 2 | Dani Pedrosa | Honda | 21 | +4.008 | 1 | 20 |\n| 3 | 4 | Andrea Dovizioso | Honda | 21 | +8.536 | 6 | 16 |\n| 4 | 69 | Nicky Hayden | Honda | 21 | +8.858 | 4 | 13 |\n| 5 | 56 | Shinya Nakano | Honda | 21 | +10.583 | 15 | 11 |\n| 6 | 1 | Casey Stoner | Ducati | 21 | +13.640 | 7 | 10 |\n| 7 | 65 | Loris Capirossi | Suzuki | 21 | +15.936 | 8 | 9 |\n| 8 | 5 | Colin Edwards | Yamaha | 21 | +18.802 | 5 | 8 |\n| 9 | 7 | Chris Vermeulen | Suzuki | 21 | +23.174 | 11 | 7 |\n| 10 | 14 | Randy de Puniet | Honda | 21 | +25.516 | 9 | 6 |\n| 11 | 21 | John Hopkins | Kawasaki | 21 | +27.609 | 10 | 5 |\n| 12 | 13 | Anthony West | Kawasaki | 21 | +41.399 | 13 | 4 |\n| 13 | 50 | Sylvain Guintoli | Ducati | 21 | +45.617 | 16 | 3 |\n| 14 | 15 | Alex de Angelis | Honda | 21 | +49.003 | 17 | 2 |\n| 15 | 24 | Toni Elias | Ducati | 21 | +59.139 | 19 | 1 |\n| 16 | 33 | Marco Melandri | Ducati | 21 | +1:03.328 | 14 | |\n| 17 | 9 | Nobuatsu Aoki | Suzuki | 21 | +1:48.363 | 18 | |\n| Ret | 48 | Jorge Lorenzo | Yamaha | 12 | Accident | 3 | |\n| Ret | 52 | James Toseland | Yamaha | 2 | Accident | 12 | |\n\n\nQuestion\ndid any country have consecutive winners in the race?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"no"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-4857",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-10472.csv` in your code execution environment.\n\nTable Title: Luis Regueiro\nTable Content:\n| # | Date | Venue | Opponent | Score | Result | Competition |\n| --- | --- | --- | --- | --- | --- | --- |\n| 1. | 30 May 1928 | Olympic Stadium (Amsterdam), Amsterdam, Netherland | Mexico | 1–0 | 7–1 | 1928 Summer Olympics |\n| 2. | 30 May 1928 | Olympic Stadium (Amsterdam), Amsterdam, Netherland | Mexico | 3–0 | 7–1 | 1928 Summer Olympics |\n| 3. | 22 June 1930 | Bologna, Italy | Italy | 1–0 | 3–2 | Friendly |\n| 4. | 22 June 1930 | Bologna, Italy | Italy | 3–1 | 3–2 | Friendly |\n| 5. | 13 December 1931 | Dublin, Ireland | Ireland | 3–0 | 5–0 | Friendly |\n| 6. | 13 December 1931 | Dublin, Ireland | Ireland | 5–0 | 5–0 | Friendly |\n| 7. | 24 April 1932 | Oviedo, Spain | Yugoslavia | 1–0 | 2–1 | Friendly |\n| 8. | 21 June 1933 | Madrid, Spain | Bulgaria | 8–0 | 13–0 | Friendly |\n| 9. | 21 June 1933 | Madrid, Spain | Bulgaria | 12–0 | 13–0 | Friendly |\n| 10. | 11 March 1934 | Madrid, Spain | Portugal | 4–0 | 9–0 | 1934 World Cup Qualifier |\n| 11. | 11 March 1934 | Madrid, Spain | Portugal | 6–0 | 9–0 | 1934 World Cup Qualifier |\n| 12. | 21 June 1934 | Firenze, Italy | Italy | 1–0 | 1–1 | 1934 FIFA World Cup |\n| 13. | 24 January 1935 | Madrid, Spain | Bulgaria | 1–0 | 2–0 | Friendly |\n| 14. | 19 January 1936 | Madrid, Spain | Austria | 2–4 | 4–5 | Friendly |\n| 15. | 19 January 1936 | Madrid, Spain | Austria | 4–4 | 4–5 | Friendly |\n| 16. | 23 February 1936 | Barcelona, Spain | Germany | 1–2 | 1–2 | Friendly |\n\n\nQuestion\nhow many consecutive 'friendly' games did spain have?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"11"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-10472",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-10931.csv` in your code execution environment.\n\nTable Title: 1960 Buffalo Bills season\nTable Content:\n| Week | Date | Opponent | Result | Game site | Attendance |\n| --- | --- | --- | --- | --- | --- |\n| 1 | September 11, 1960 | at New York Titans | L 27–3 | Polo Grounds | 10,200 |\n| 2 | September 18, 1960 | Denver Broncos | L 27–21 | War Memorial Stadium | 15,229 |\n| 3 | September 23, 1960 | at Boston Patriots | W 13–0 | Nickerson Field | 20,732 |\n| 4 | October 2, 1960 | Los Angeles Chargers | L 24–10 | War Memorial Stadium | 15,821 |\n| 6 | October 16, 1960 | New York Titans | L 17–13 | War Memorial Stadium | 14,988 |\n| 7 | October 23, 1960 | Oakland Raiders | W 38–9 | War Memorial Stadium | 8,876 |\n| 8 | October 30, 1960 | Houston Oilers | W 25–24 | War Memorial Stadium | 23,001 |\n| 9 | November 6, 1960 | Dallas Texans | L 45–28 | War Memorial Stadium | 19,610 |\n| 10 | November 13, 1960 | at Oakland Raiders | L 20–7 | Kezar Stadium | 8,800 |\n| 11 | November 20, 1960 | at Los Angeles Chargers | W 32–3 | Los Angeles Memorial Coliseum | 16,161 |\n| 12 | November 27, 1960 | at Denver Broncos | T 38–38 | Bears Stadium | 7,785 |\n| 13 | December 4, 1960 | Boston Patriots | W 38–14 | War Memorial Stadium | 14,335 |\n| 14 | December 11, 1960 | at Houston Oilers | L 31–23 | Jeppesen Stadium | 25,243 |\n| 15 | December 18, 1960 | at Dallas Texans | L 24–7 | Cotton Bowl | 18,000 |\n\n\nQuestion\nhow many games had at least 10,000 people in attendance?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"11"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-10931",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-9035.csv` in your code execution environment.\n\nTable Title: 1961–62 Segunda División\nTable Content:\n| Position | Club | Played | Points | Wins | Draws | Losses | Goals for | Goals against | Goal Difference |\n| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |\n| 1 | Córdoba CF | 30 | 40 | 16 | 8 | 6 | 48 | 22 | +26 |\n| 2 | CD Málaga | 30 | 38 | 14 | 10 | 6 | 52 | 36 | +16 |\n| 3 | Granada CF | 30 | 36 | 15 | 6 | 9 | 48 | 34 | +14 |\n| 4 | UD Las Palmas | 30 | 35 | 15 | 5 | 10 | 47 | 39 | +8 |\n| 5 | Recreativo de Huelva | 30 | 33 | 13 | 7 | 10 | 43 | 42 | +1 |\n| 6 | Levante UD | 30 | 32 | 14 | 4 | 12 | 49 | 42 | +7 |\n| 7 | Hércules CF | 30 | 32 | 14 | 4 | 12 | 55 | 46 | +9 |\n| 8 | Real Murcia | 30 | 31 | 12 | 7 | 11 | 40 | 35 | +5 |\n| 9 | Real Jaén | 30 | 31 | 14 | 3 | 13 | 58 | 42 | +16 |\n| 10 | Cádiz CF | 30 | 28 | 12 | 4 | 14 | 43 | 52 | -9 |\n| 11 | CD Cartagena | 30 | 28 | 13 | 2 | 15 | 45 | 56 | -11 |\n| 12 | CD Mestalla | 30 | 27 | 11 | 5 | 14 | 50 | 49 | +1 |\n| 13 | Albacete Balompié | 30 | 27 | 10 | 7 | 13 | 27 | 32 | -5 |\n| 14 | CD San Fernando | 30 | 27 | 11 | 5 | 14 | 37 | 47 | -10 |\n| 15 | Atlético Ceuta | 30 | 23 | 8 | 7 | 15 | 33 | 48 | -15 |\n| 16 | CD Villarrobledo | 30 | 12 | 4 | 4 | 22 | 26 | 79 | -53 |\n\n\nQuestion\nwhat were the number of goals for listed for the hercules cf club?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"55"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-9035",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-1104.csv` in your code execution environment.\n\nTable Title: 1988–89 FA Cup\nTable Content:\n| Tie no | Home team | Score | Away team | Date |\n| --- | --- | --- | --- | --- |\n| 1 | Enfield | 1–1 | Leyton Orient | 19 November 1988 |\n| Replay | Leyton Orient | 2–2 | Enfield | 23 November 1988 |\n| Replay | Leyton Orient | 0–1 | Enfield | 28 November 1988 |\n| 2 | Blackpool | 2–1 | Scunthorpe United | 19 November 1988 |\n| 3 | Darlington | 1–2 | Notts County | 19 November 1988 |\n| 4 | Bath City | 2–0 | Grays Athletic | 19 November 1988 |\n| 5 | Bristol City | 3–1 | Southend United | 19 November 1988 |\n| 6 | Burnley | 0–2 | Chester City | 19 November 1988 |\n| 7 | Preston North End | 1–1 | Tranmere Rovers | 19 November 1988 |\n| Replay | Tranmere Rovers | 3–0 | Preston North End | 22 November 1988 |\n| 8 | Yeovil Town | 3–2 | Merthyr Tydfil | 19 November 1988 |\n| 9 | Reading | 4–2 | Hendon | 19 November 1988 |\n| 10 | Woking | 1–4 | Cambridge United | 19 November 1988 |\n| 11 | Gillingham | 3–3 | Peterborough United | 19 November 1988 |\n| Replay | Peterborough United | 1–0 | Gillingham | 23 November 1988 |\n| 12 | Bolton Wanderers | 0–0 | Chesterfield | 19 November 1988 |\n| Replay | Chesterfield | 2–3 | Bolton Wanderers | 28 November 1988 |\n| 13 | Grimsby Town | 1–0 | Wolverhampton Wanderers | 19 November 1988 |\n| 14 | Stafford Rangers | 2–2 | Crewe Alexandra | 19 November 1988 |\n| Replay | Crewe Alexandra | 3–2 | Stafford Rangers | 22 November 1988 |\n| 15 | Scarborough | 2–1 | Stockport County | 19 November 1988 |\n| 16 | Doncaster Rovers | 0–0 | Brandon United | 19 November 1988 |\n| Replay | Doncaster Rovers | 2–1 | Brandon United | 22 November 1988 |\n| 17 | Fulham | 0–1 | Colchester United | 19 November 1988 |\n| 18 | Brentford | 2–0 | Halesowen Town | 19 November 1988 |\n| 19 | Bristol Rovers | 3–0 | Fisher Athletic | 20 November 1988 |\n| 20 | Altrincham | 3–2 | Lincoln City | 19 November 1988 |\n| 21 | Bognor Regis Town | 2–1 | Exeter City | 19 November 1988 |\n| 22 | Huddersfield Town | 1–1 | Rochdale | 19 November 1988 |\n| Replay | Rochdale | 3–4 | Huddersfield Town | 28 November 1988 |\n| 23 | Mansfield Town | 1–1 | Sheffield United | 19 November 1988 |\n| Replay | Sheffield United | 2–1 | Mansfield Town | 22 November 1988 |\n| 24 | Cardiff City | 3–0 | Hereford United | 19 November 1988 |\n| 25 | Halifax Town | 1–0 | York City | 19 November 1988 |\n| 26 | Newport County | 1–2 | Maidstone United | 19 November 1988 |\n| 27 | Southport | 0–2 | Port Vale | 19 November 1988 |\n| 28 | Runcorn | 2–2 | Wrexham | 19 November 1988 |\n| Replay | Wrexham | 2–3 | Runcorn | 22 November 1988 |\n| 29 | Torquay United | 2–2 | Fareham Town | 19 November 1988 |\n| Replay | Fareham Town | 2–3 | Torquay United | 23 November 1988 |\n| 30 | Kettering Town | 2–1 | Dartford | 19 November 1988 |\n| 31 | Rotherham United | 3–1 | Barrow | 19 November 1988 |\n| 32 | Aldershot | 1–0 | Hayes | 19 November 1988 |\n| 33 | Dagenham | 0–4 | Sutton United | 19 November 1988 |\n| 34 | Waterlooville | 1–4 | Aylesbury United | 19 November 1988 |\n| 35 | Telford United | 1–1 | Carlisle United | 19 November 1988 |\n| Replay | Carlisle United | 4–1 | Telford United | 22 November 1988 |\n| 36 | Swansea City | 3–1 | Northampton Town | 19 November 1988 |\n| 37 | Frickley Athletic | 0–2 | Northwich Victoria | 19 November 1988 |\n| 38 | Hartlepool United | 2–0 | Wigan Athletic | 19 November 1988 |\n| 39 | Welling United | 3–0 | Bromsgrove Rovers | 19 November 1988 |\n| 40 | Guisborough Town | 0–1 | Bury | 19 November 1988 |\n\n\nQuestion\nwhat is the number of games that gillingham and peterborough united played in the first round proper?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"2"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-1104",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-2898.csv` in your code execution environment.\n\nTable Title: Bishop of Iceland\nTable Content:\n| # | Name | Birth and death | Office started | Office ended |\n| --- | --- | --- | --- | --- |\n| 1 | Geir Vídalín | 27 October 1761\\n–\\n20 September 1823 | 1801 | 1823 |\n| 2 | Steingrímur Jónsson | | 1824 | 1845 |\n| 3 | Helgi Thordersen | 8 April 1794\\n–\\n4 December 1867 | 1846 | 1866 |\n| 4 | 'Pétur Pétursson | 3 October 1808\\n–\\n15 May 1891 | 1866 | 1889 |\n| 5 | Hallgrímur Sveinsson | 5 April 1841\\n–\\n16 December 1909 | 1889 | 1908 |\n| 6 | Þórhallur Bjarnarson | 2 December 1855\\n–\\n15 December 1916 | 1908 | 1916 |\n| 7 | Jón Helgason | 1866\\n–\\n1942 | 1917 | 1939 |\n| 8 | Sigurgeir Sigurðsson | 3 August 1890\\n-\\n13 October 1953 | 1939 | 1953 |\n| 9 | Ásmundur Guðmundsson | 6 October 1888\\nReykholt\\n–\\n29 May 1969\\nReykjavík | 1954 | 1989 |\n| 10 | Sigurbjörn Einarsson | 30 June 1911\\nVestur-Skaftafellssýsla\\n–\\n28 August 2008\\nReykjavík | 1959 | 1981 |\n| 11 | Pétur Sigurgeirsson\\n(son of Sigurgeir Sigurðsson, 8th Bishop of Iceland) | 2 June 1919\\n–\\n3 June 2010 | 1981 | 1989 |\n| 12 | Ólafur Skúlason | 29 December 1929\\n–\\n9 June 2008 | 1989 | 1997 |\n| 13 | Karl Sigurbjörnsson\\n(son of Sigurbjörn Einarsson, 10th Bishop of Iceland) | 5 February 1947\\nReykjavík | 1998 | 2012 |\n| 14 | Agnes Sigurðardóttir | 19 October 1954\\nÍsafjörður | 24 June 2012 | Incumbent |\n\n\nQuestion\nwhich bishop served between the years of 1846 and 1866?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"Helgi Thordersen"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-2898",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-6985.csv` in your code execution environment.\n\nTable Title: List of National Natural Landmarks in Hawaii\nTable Content:\n| | Name | Date | Location | County | Description |\n| --- | --- | --- | --- | --- | --- |\n| 1 | Diamond Head | | Honolulu | Oahu | State Monument |\n| 2 | I'ao Valley | | Wailuku | Maui | State Monument |\n| 3 | Kanaha Pond | June 1971 | Kahului | Maui | Bird Sanctuary |\n| 4 | Koʻolau Range Pali | | | Oahu | |\n| 5 | Makalawena Marsh | June 1972 | North Kona | Hawaiʻi | Bird Sanctuary |\n| 6 | Mauna Kea | November 1972 | Saddle Road | Hawaiʻi | U.S. state high point |\n| 7 | North Shore Cliffs | December 1972 | Molokaʻi | Maui | |\n\n\nQuestion\nother than mauna kea, name a place in hawaii.\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"Makalawena Marsh"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-6985",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-9789.csv` in your code execution environment.\n\nTable Title: 2002–03 Edmonton Oilers season\nTable Content:\n| Round | # | Player | Nationality | College/Junior/Club Team (League) |\n| --- | --- | --- | --- | --- |\n| 1 | 15 | Jesse Niinimäki | Finland | Ilves (Finland) |\n| 2 | 31 | Jeff Deslauriers | Canada | Chicoutimi Saguenéens (QMJHL) |\n| 2 | 36 | Jarret Stoll | Canada | Kootenay Ice (WHL) |\n| 2 | 44 | Matt Greene | United States | Green Bay Gamblers (USHL) |\n| 3 | 79 | Brock Radunske | Canada | Michigan State Spartans (NCAA) |\n| 4 | 106 | Ivan Koltsov | Russia | Cherepovets (Russia) |\n| 4 | 111 | Jonas Almtorp | Sweden | Modo Hockey (Sweden) |\n| 4 | 123 | Robin Kovář | Czech Republic | HC Vsetin (Czech.) |\n| 5 | 148 | Glenn Fisher | Canada | Fort Saskatchewan Traders (AJHL) |\n| 6 | 181 | Mikko Luoma | Finland | Tappara (Finland) |\n| 7 | 205 | Jean-Francois Dufort | Canada | Cape Breton Screaming Eagles (QMJHL) |\n| 7 | 211 | Patrick Murphy | Canada | Newmarket Hurricanes (OPJHL) |\n| 8 | 244 | Dwight Helminen | United States | Michigan Wolverines (NCAA) |\n| 8 | 245 | Tomáš Micka | Czech Republic | HC Slavia Praha (Czech.) |\n| 9 | 274 | Fredrik Johansson | Sweden | Frolunda (Sweden) |\n\n\nQuestion\nlist each of the players from finland.\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"Jesse Niinimäki",
"Mikko Luoma"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-9789",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-10841.csv` in your code execution environment.\n\nTable Title: 1971 World Figure Skating Championships\nTable Content:\n| Rank | Name | Nation | CP | FP | Points | Placings |\n| --- | --- | --- | --- | --- | --- | --- |\n| 1 | Beatrix Schuba | Austria | 1 | 7 | 2763.0 | 10 |\n| 2 | Julie Lynn Holmes | United States | 2 | 5 | 2697.3 | 24.5 |\n| 3 | Karen Magnussen | Canada | 4 | 2 | 2697.7 | 27 |\n| 4 | Janet Lynn | United States | 5 | 1 | 2680.0 | 34 |\n| 5 | Rita Trapanese | Italy | 3 | 8 | 2623.6 | 47.5 |\n| 6 | Sonja Morgenstern | East Germany | 9 | 3 | 2606.6 | 52 |\n| 7 | Zsuzsa Almássy | Hungary | 6 | 4 | 2599.9 | 57 |\n| 8 | Charlotte Walter | Switzerland | 8 | 14 | 2480.7 | 86 |\n| 9 | Christine Errath | East Germany | 13 | 6 | 2493.5 | 83 |\n| 10 | Suna Murray | United States | 14 | 9 | 2475.0 | 91 |\n| 11 | Elena Alexandrova | Soviet Union | 10 | 13 | 2450.5 | 103 |\n| 12 | Patricia Ann Dodd | United Kingdom | 7 | 16 | | 100 |\n| 13 | Kazumi Yamashita | Japan | 12 | 11 | | 116 |\n| 14 | Ludmila Bezáková | Czechoslovakia | 11 | 15 | | 122 |\n| 15 | Jean Scott | United Kingdom | 15 | 10 | | 132 |\n| 16 | Diane Hall | Canada | 19 | 12 | | 132 |\n| 17 | Anita Johansson | Sweden | 16 | 18 | | 145 |\n| 18 | Judith Bayer | West Germany | 20 | 17 | | 159 |\n| 19 | Ruth Hutchinson | Canada | 17 | 20 | | 168 |\n| 20 | Sonja Balun | Austria | 22 | 19 | | 177 |\n| 21 | Joëlle Cartaux | France | 18 | 22 | | 188 |\n| 22 | Cinzia Frosio | Italy | 21 | 21 | | 197 |\n\n\nQuestion\nwhich woman finished last in the 1971 world figured skating championships?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"Cinzia Frosio"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-10841",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-5753.csv` in your code execution environment.\n\nTable Title: 2003 Asian Athletics Championships – Women's triple jump\nTable Content:\n| Rank | Name | Nationality | Result | Notes |\n| --- | --- | --- | --- | --- |\n| | Huang Qiuyan | China | 14.39 | SB |\n| | Anastasiya Juravleva | Uzbekistan | 14.21 | =SB |\n| | Zhang Hao | China | 13.63 | |\n| 4 | Wacharee Ritthiwat | Thailand | 13.54 | SB |\n| 5 | Tatyana Bocharova | Kazakhstan | 13.50 | |\n| 6 | Wang Kuo-Huei | Chinese Taipei | 13.25 | SB |\n| 7 | Fumiyo Yoshida | Japan | 13.09 | |\n| 8 | Ngew Sin Mei | Malaysia | 12.93 | |\n| 9 | Manisha Dey | India | 12.83 | |\n| 10 | Thitima Muangjan | Thailand | 12.79 | |\n| 11 | Phan Thi Thu Lan | Vietnam | 12.75 | |\n| 12 | Lo Yu-Hsin | Chinese Taipei | 12.69 | |\n| | Elena Bobrovskaya | Kyrgyzstan | DNS | |\n\n\nQuestion\nhow many points apart were the 1st place competitor and the 12th place competitor?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"1.7"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-5753",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-8219.csv` in your code execution environment.\n\nTable Title: 2013 Vattenfall Cyclassics\nTable Content:\n| | Cyclist | Team | Time | UCI World Tour\\nPoints |\n| --- | --- | --- | --- | --- |\n| 1 | John Degenkolb (GER) | Argos-Shimano | 5h 45' 16\" | 80 |\n| 2 | André Greipel (GER) | Lotto-Belisol | s.t. | 60 |\n| 3 | Alexander Kristoff (NOR) | Team Katusha | + 1\" | 50 |\n| 4 | José Joaquín Rojas (ESP) | Movistar Team | + 1\" | 40 |\n| 5 | Elia Viviani (ITA) | Cannondale | + 1\" | 30 |\n| 6 | Boy Van Poppel (NLD) | Vacansoleil-DCM | + 1\" | 22 |\n| 7 | Nikolas Maes (BEL) | Omega Pharma-Quick Step | + 1\" | 14 |\n| 8 | Thor Hushovd (NOR) | BMC Racing Team | + 1\" | 10 |\n| 9 | Matti Breschel (DNK) | Team Saxo-Tinkoff | + 1\" | 6 |\n| 10 | Arnaud Démare (FRA) | FDJ.fr | + 1\" | 2 |\n\n\nQuestion\nwas josé joaquín rojas from the movistar team or team katusha?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"Movistar Team"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-8219",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-7132.csv` in your code execution environment.\n\nTable Title: Biathlon World Championships 2013 – Men's pursuit\nTable Content:\n| Rank | Bib | Name | Nationality | Start | Penalties (P+P+S+S) | Time | Deficit |\n| --- | --- | --- | --- | --- | --- | --- | --- |\n| | 1 | Emil Hegle Svendsen | Norway | 0:00 | 1 (0+0+0+1) | 32:35.5 | |\n| | 2 | Martin Fourcade | France | 0:08 | 2 (0+1+1+0) | 32:35.6 | +0.1 |\n| | 7 | Anton Shipulin | Russia | 0:33 | 1 (0+0+1+0) | 32:39.1 | +3.6 |\n| 4 | 5 | Dmitry Malyshko | Russia | 0:23 | 0 (0+0+0+0) | 32:40.9 | +5.4 |\n| 5 | 15 | Dominik Landertinger | Austria | 0:58 | 0 (0+0+0+0) | 32:55.9 | +20.4 |\n| 6 | 3 | Jakov Fak | Slovenia | 0:11 | 1 (0+0+0+1) | 33:05.2 | +29.7 |\n| 7 | 8 | Fredrik Lindström | Sweden | 0:38 | 1 (0+0+1+0) | 33:18.9 | +43.4 |\n| 8 | 6 | Alexis Bœuf | France | 0:25 | 2 (0+0+2+0) | 33:22.1 | +46.6 |\n| 9 | 13 | Björn Ferry | Sweden | 0:52 | 0 (0+0+0+0) | 33:23.6 | +48.1 |\n| 10 | 4 | Ole Einar Bjørndalen | Norway | 0:20 | 4 (2+0+1+1) | 33:27.0 | +51.5 |\n| 11 | 14 | Lukas Hofer | Italy | 0:57 | 2 (0+2+0+0) | 33:28.8 | +53.3 |\n| 12 | 10 | Simon Eder | Austria | 0:39 | 3 (0+0+2+1) | 33:50.5 | +1:15.0 |\n| 13 | 32 | Lowell Bailey | United States | 1:38 | 0 (0+0+0+0) | 33:56.4 | +1:20.9 |\n| 14 | 12 | Erik Lesser | Germany | 0:49 | 4 (1+0+2+1) | 34:06.0 | +1:30.5 |\n| 15 | 36 | Jean-Philippe Leguellec | Canada | 1:45 | 1 (0+0+0+1) | 34:08.7 | +1:33.2 |\n| 16 | 9 | Evgeny Ustyugov | Russia | 0:38 | 3 (0+1+2+0) | 34:16.7 | +1:41.2 |\n| 17 | 18 | Tarjei Bø | Norway | 1:14 | 3 (2+0+1+0) | 34:18.9 | +1:43.4 |\n| 18 | 29 | Simon Schempp | Germany | 1:36 | 1 (0+0+0+1) | 34:24.4 | +1:48.9 |\n| 19 | 11 | Krasimir Anev | Bulgaria | 0:48 | 2 (0+1+0+1) | 34:24.7 | +1:49.2 |\n| 20 | 37 | Henrik L'Abee-Lund | Norway | 1:45 | 1 (0+0+0+1) | 34:36.1 | +2:00.6 |\n| 21 | 16 | Arnd Peiffer | Germany | 1:08 | 3 (0+0+2+1) | 34:39.8 | +2:04.3 |\n| 22 | 23 | Andreas Birnbacher | Germany | 1:25 | 3 (0+1+1+1) | 34:41.7 | +2:06.2 |\n| 23 | 34 | Simon Fourcade | France | 1:42 | 3 (0+3+0+0) | 34:51.1 | +2:15.6 |\n| 24 | 21 | Klemen Bauer | Slovenia | 1:16 | 4 (1+0+3+0) | 34:55.4 | +2:19.9 |\n| 25 | 26 | Christian De Lorenzi | Italy | 1:30 | 1 (0+0+0+1) | 34:57.5 | +2:22.0 |\n| 26 | 20 | Jean-Guillaume Béatrix | France | 1:15 | 4 (0+1+2+1) | 34:58.4 | +2:22.9 |\n| 27 | 33 | Daniel Mesotitsch | Austria | 1:39 | 3 (0+0+2+1) | 34:58.7 | +2:23.2 |\n| 28 | 19 | Evgeniy Garanichev | Russia | 1:14 | 5 (2+0+2+1) | 35:04.2 | +2:28.7 |\n| 29 | 30 | Matej Kazár | Slovakia | 1:36 | 5 (2+1+1+1) | 35:04.9 | +2:29.4 |\n| 30 | 17 | Andriy Deryzemlya | Ukraine | 1:13 | 4 (0+0+2+2) | 35:05.3 | +2:29.8 |\n| 31 | 35 | Ondřej Moravec | Czech Republic | 1:42 | 3 (0+1+1+1) | 35:05.3 | +2:48.2 |\n| 32 | 28 | Tim Burke | United States | 1:36 | 4 (0+1+2+1) | 35:33.8 | +2:58.3 |\n| 33 | 58 | Dominik Windisch | Italy | 2:31 | 2 (0+0+1+1) | 35:36.4 | +3:00.9 |\n| 34 | 54 | Carl Johan Bergman | Sweden | 2:19 | 2 (0+0+1+1) | 35:37.0 | +3:01.5 |\n| 35 | 25 | Benjamin Weger | Switzerland | 1:28 | 3 (0+1+1+1) | 36:13.1 | +3:37.6 |\n| 36 | 42 | Artem Pryma | Ukraine | 1:59 | 3 (1+2+0+0) | 36:13.2 | +3:37.7 |\n| 37 | 50 | Pavol Hurajt | Slovakia | 2:14 | 2 (0+1+0+1) | 36:18.9 | +3:43.4 |\n| 38 | 22 | Tomas Kaukėnas | Lithuania | 1:20 | 4 (0+0+2+2) | 36:33.4 | +3:57.9 |\n| 39 | 43 | Vladimir Iliev | Bulgaria | 2:03 | 5 (1+1+2+1) | 36:43.4 | +4:07.9 |\n| 40 | 40 | Edgars Piksons | Latvia | 1:55 | 4 (1+0+2+1) | 36:49.8 | +4:14.3 |\n| 41 | 60 | Christian Martinelli | Italy | 2:36 | 2 (1+0+0+1) | 36:53.0 | +4:17.5 |\n| 42 | 27 | Jaroslav Soukup | Czech Republic | 1:32 | 6 (2+2+2+0) | 36:59.8 | +4:24.3 |\n| 43 | 53 | Leif Nordgren | United States | 2:19 | 6 (2+2+2+0) | 37:08.3 | +4:32.8 |\n| 44 | 56 | Hidenori Isa | Japan | 2:27 | 2 (0+0+0+2) | 37:13.9 | +4:38.4 |\n| 45 | 41 | Serhiy Semenov | Ukraine | 1:55 | 5 (3+1+1+0) | 37:14.0 | +4:38.5 |\n| 46 | 24 | Zdeněk Vítek | Czech Republic | 1:25 | 7 (2+2+1+2) | 37:21.0 | +4:45.5 |\n| 47 | 49 | Milanko Petrović | Serbia | 2:14 | 6 (2+2+2+0) | 37:26.6 | +4:51.1 |\n| 48 | 31 | Kauri Kõiv | Estonia | 1:37 | 6 (2+1+2+1) | 37:29.9 | +4:54.4 |\n| 49 | 47 | Yan Savitskiy | Kazakhstan | 2:11 | 5 (2+2+0+1) | 37:51.8 | +5:16.3 |\n| 50 | 45 | Scott Gow | Canada | 2:06 | 5 (0+0+1+4) | 37:57.4 | +5:21.9 |\n| 51 | 48 | Claudio Böckli | Switzerland | 2:13 | 3 (2+0+0+1) | 38:02.8 | +5:27.3 |\n| 52 | 44 | Russell Currier | United States | 2:05 | 7 (3+1+2+1) | 38:06.2 | +5:30.7 |\n| 53 | 57 | Vladimir Chepelin | Belarus | 2:28 | 4 (0+1+2+1) | 38:06.5 | +5:31.0 |\n| 54 | 46 | Aliaksandr Babchyn | Belarus | 2:06 | 6 (3+2+1+0) | 38:19.8 | +5:44.3 |\n| 55 | 59 | Miroslav Matiaško | Slovakia | 2:36 | 6 (3+0+2+1) | 38:54.6 | +6:19.1 |\n| 56 | 38 | Jarkko Kauppinen | Finland | 1:46 | 6 (1+1+3+1) | 39:06.6 | +6:31.1 |\n| 57 | 51 | Junji Nagai | Japan | 2:17 | 7 (0+2+3+2) | 39:49.9 | +7:14.4 |\n| | 55 | Stefan Gavrila | Romania | 2:25 | 10 (4+3+3) | LAP | |\n| | 39 | Sergei Sednev | Ukraine | 1:54 | | DNS | |\n| | 52 | Michal Šlesingr | Czech Republic | 2:18 | | DNS | |\n\n\nQuestion\nwhat were the total number of \"ties\" (people who finished with the exact same time?)\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"2"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-7132",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-12827.csv` in your code execution environment.\n\nTable Title: Clint Newton\nTable Content:\n| Date | Venue | Opponent | Result | Tournament | Scored |\n| --- | --- | --- | --- | --- | --- |\n| 10 February 2008 | Craven Park | St. Helens | 24–22 | Super League XIII | 1 |\n| 24 March 2008 | JJB Stadium | Wigan Warriors | 18–12 | Super League XIII | 1 |\n| 13 April 2008 | Craven Park | Bradford Bulls | 20–18 | Super League XIII | 1 |\n| 21 June 2008 | Craven Park | Wakefield Trinity Wildcats | 18–26 | Super League XIII | 1 |\n| 8 March 2009 | Craven Park | Wakefield Trinity Wildcats | 31–18 | Super League XIV | 1 |\n| 15 March 2009 | Craven Park | Crusaders | 48–18 | Super League XIV | 3 |\n| 3 May 2009 | Murrayfield Stadium | Warrington Wolves | 28–36 | Super League XIV | 1 |\n| 20 June 2009 | Brewery Field | Crusaders | 18–32 | Super League XIV | 2 |\n| 21 August 2009 | The Willows | Salford City Reds | 10–14 | Super League XIV | 1 |\n| 5 September 2009 | Craven Park | Warrington Wolves | 40–16 | Super League XIV | 3 |\n| 12 September 2009 | Belle Vue | Wakefield Trinity Wildcats | 24–10 | Super League XIV | 1 |\n| 7 February 2010 | Craven Park | Salford City Reds | 30–12 | Super League XV | 1 |\n| 16 May 2010 | Craven Park | Crusaders | 54-10 | Super League XV | 2 |\n| 21 May 2010 | Knowsley Road | St. Helens | 68-12 | Super League XV | 1 |\n| 20 June 2010 | Craven Park | Harlequins RL | 42-6 | Super League XV | 2 |\n| 25 June 2010 | Galpharm Stadium | Huddersfield Giants | 16-16 | Super League XV | 1 |\n| 4 July 2010 | Belle Vue | Wakefield Trinity Wildcats | 14-46 | Super League XV | 1 |\n| 15 July 2010 | KC Stadium | Hull | 20-16 | Super League XV | 1 |\n| 24 July 2010 | Craven Park | Bradford Bulls | 49-24 | Super League XV | 2 |\n| 15 August 2010 | The Willows | Salford City Reds | 18-44 | Super League XV | 2 |\n| 13 February 2011 | Millennium Stadium | Hull | 22-34 | Super League XV | 1 |\n| 20 February 2011 | Halliwell Jones Stadium | Warrington Wolves | 24-22 | Super League | 2 |\n| 7 March 2011 | The Jungle | Castleford Tigers | 27-14 | Super League | 1 |\n| 25 June 2011 | Twickenham Stoop | Harlequins RL | 0-34 | Super League | 1 |\n\n\nQuestion\nname a match that had a total score above 40 but less than 65.\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"10 February 2008"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-12827",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-12697.csv` in your code execution environment.\n\nTable Title: Pericardial heart valves\nTable Content:\n| Main author | No. of patients | Duration of follow-up years | Linearized rate of infection | Actuarial freedom from infective endocarditis |\n| --- | --- | --- | --- | --- |\n| Pavie | 675 | 5 | | 98.2% for all patients, 97.8% for aortic, 99% for mitral, 100% for multiple valve replacements |\n| Duncan (A) | 2720 | 6 | | 97.3% for all patients, 97.4% for aortic valves, 97.6% for mitral valves, 96.3% for multiple valve replacement |\n| Ionescu (B) | 1171 | 10 | | 93.7% for all patients, 94.7% for aortic valves, 97.1% for mitral valves, 89.3% for multiple valve replacement |\n| Zhu | 520 | 10 | | 98% for all patients |\n| Revuelta | 239 | 8 | 0.67% per patient year | |\n| Garcia-Bengochea | 248 | 8 | 0.78% per patient year | |\n| Holden (C) | 290 | 6 | 1.1% per patient year | |\n| Bachet (D) | 224 | 6 | 1.6% per patient year | |\n\n\nQuestion\nwhat is the minimum of follow-up years do each author have?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"5"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-12697",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-13759.csv` in your code execution environment.\n\nTable Title: Didier Ya Konan\nTable Content:\n| # | Date | Venue | Opponent | Score | Result | Competition |\n| --- | --- | --- | --- | --- | --- | --- |\n| 1 | 8 February 2011 | Stade Georges Pompidou, Valence, France | Mali | 1–0 | 1–0 | Friendly |\n| 2 | 11 May 2011 | Stade Félix Houphouët-Boigny, Abidjan, Côte d'Ivoire | Namibia | 4–1 | 5–1 | Friendly |\n| 3 | 5 June 2011 | Stade de l'Amitié, Cotonou, Benin | Benin | 0–1 | 2–6 | 2012 Africa Cup of Nations qualification – Group H |\n| 4 | 10 August 2011 | Stade de Geneve, Geneve, Switzerland | Israel | 1–0 | 4–3 | Friendly |\n| 5 | 3 September 2011 | Stade Amahoro, Kigali, Rwanda | Rwanda | 0–4 | 0–5 | 2012 Africa Cup of Nations qualification – Group H |\n| 6 | 14 November 2012 | Linzer Stadion, Linz, Austria | Austria | 0–1 | 0–3 | Friendly |\n| 7 | 14 January 2013 | Al Nahyan Stadium, Abu Dhabi, United Arab Emirates | Egypt | 4–2 | 4–2 | Friendly |\n| 8 | 26 January 2013 | Royal Bafokeng Stadium, Rustenburg, South Africa | Tunisia | 3–0 | 3–0 | 2013 Africa Cup of Nations |\n\n\nQuestion\nwhich competition took place after the game in egypt?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"2013 Africa Cup of Nations"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-13759",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-7085.csv` in your code execution environment.\n\nTable Title: Judges of the Permanent Court of International Justice\nTable Content:\n| Nationality | Name | Term as a Deputy Judge | Reason for termination |\n| --- | --- | --- | --- |\n| Norway | Frederik Beichmann | 30 January 1920 — 6 December 1930 | Not re-elected |\n| China | Wang Ch'ung-hui | 30 January 1920 — 6 December 1930 | Not re-elected |\n| Finland | Rafael Erich | 15 January 1931 — 1 February 1936 | Post abolished |\n| Portugal | Jose Caeiro da Matta | 15 January 1931 — 1 February 1936 | Post abolished |\n| Romania | Demetre Negulesco | 30 January 1920 — 6 December 1930 | Not re-elected |\n| Yugoslavia | Mileta Novaković | 15 January 1931 — 1 February 1936 | Post abolished |\n| Austria | Josef Redlich | 15 January 1931 — 1 February 1936 | Post abolished |\n| Yugoslavia | Mihajlo Jovanović | 30 January 1920 — 6 December 1930 | Not re-elected |\n\n\nQuestion\nwho was the last deputy judge with finland nationality?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"Rafael Erich"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-7085",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-13429.csv` in your code execution environment.\n\nTable Title: 1993–94 NHL season\nTable Content:\n| Date | Winning Team | Score | Losing Team | Score | OT | City | State/Province | Arena | Attendance |\n| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |\n| October 21, 1993 | St. Louis | 5 | San Jose | 2 | | Sacramento | CA | ARCO Arena | 7,144 |\n| October 31, 1993 | NY Rangers | 4 | New Jersey | 1 | | Halifax | NS | Halifax Metro Centre | 8,200 |\n| November 3, 1993 | Pittsburgh | 6 | Buffalo | 2 | | Sacramento | CA | ARCO Arena | 10,117 |\n| November 9, 1993 | Anaheim | 4 | Dallas | 2 | | Phoenix | AZ | America West Arena | 8,143 |\n| November 18, 1993 | NY Islanders | 5 | Montréal | 1 | | Hamilton | ON | Copps Coliseum | 17,008 |\n| December 9, 1993 | Dallas | 6 | Ottawa | 1 | | Minneapolis | MN | Target Center | 14,058 |\n| December 23, 1993 | Vancouver | 4 | Calgary | 3 | | Saskatoon | SK | SaskPlace | 11,429* |\n| December 31, 1993 | Philadelphia | 4 | Boston | 3 | | Minneapolis | MN | Target Center | 10,855 |\n| January 4, 1994 | Tampa Bay | 1 | Toronto | 0 | | Hamilton | ON | Copps Coliseum | 17,526* |\n| January 5, 1994 | Montréal | 2 | Québec | 0 | | Phoenix | AZ | America West Arena | 11,393 |\n| January 6, 1994 | St. Louis | 2 | Hartford | 1 | | Cleveland | OH | Richfield Coliseum | 6,956 |\n| January 17, 1994 | Detroit | 6 | Tampa Bay | 3 | | Minneapolis | MN | Target Center | 8,764 |\n| January 23, 1994 | Vancouver | 5 | Edmonton | 4 | (OT) | Saskatoon | SK | SaskPlace | N/A |\n| January 24, 1994 | Calgary | 3 | Los Angeles | 3 | (OT) | Phoenix | AZ | America West Arena | 14,864 |\n| February 2, 1994 | Washington | 5 | Philadelphia | 2 | | Cleveland | OH | Richfield Coliseum | 8,312 |\n| February 8, 1994 | San Jose | 4 | Chicago | 3 | | Sacramento | CA | ARCO Arena | 14,182* |\n| February 22, 1994 | Florida | 3 | Winnipeg | 2 | | Hamilton | ON | Copps Coliseum | 6,291 |\n| February 24, 1994 | Detroit | 3 | Hartford | 0 | | Cleveland | OH | Richfield Coliseum | 11,621 |\n| March 4, 1994 | Winnipeg | 6 | Ottawa | 1 | | Minneapolis | MN | Target Center | 6,388 |\n| March 8, 1994 | Chicago | 3 | Anaheim | 0 | | Phoenix | AZ | America West Arena | 13,847 |\n| March 9, 1994 | NY Rangers | 7 | Washington | 5 | | Halifax | NS | Halifax Metro Centre | 9,200* |\n| March 18, 1994 | Buffalo | 2 | NY Islanders | 2 | (OT) | Minneapolis | MN | Target Center | 8,016 |\n| March 23, 1994 | Florida | 1 | Toronto | 1 | (OT) | Hamilton | ON | Copps Coliseum | 17,096* |\n| March 27, 1994 | New Jersey | 5 | Quebec | 2 | | Minneapolis | MN | Target Center | 6,222 |\n| April 3, 1994 | Pittsburgh | 6 | Boston | 2 | | Cleveland | OH | Richfield Coliseum | 17,224 |\n| April 3, 1994 | Los Angeles | 6 | Edmonton | 1 | | Sacramento | CA | ARCO Arena | 10,363 |\n\n\nQuestion\ndid dallas or ottawa win the december 9, 1993 game?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"Dallas"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-13429",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-6141.csv` in your code execution environment.\n\nTable Title: Siham Hilali\nTable Content:\n| Year | Competition | Venue | Position | Event | Notes |\n| --- | --- | --- | --- | --- | --- |\n| 2003 | World Youth Championships | Sherbrooke, Canada | 1st | 3000 m | 9:12.70 |\n| 2004 | World Junior Championships | Grosseto, Italy | 3rd | 1500 m | 4:17.39 |\n| 2004 | World Junior Championships | Grosseto, Italy | 3rd | 3000 m | 9:03.16 (PB) |\n| 2007 | World Athletics Final | Stuttgart, Germany | 8th | 1500 m | 4:16.51 |\n| 2007 | World Championships | Osaka, Japan | 31st (h) | 1500 m | 4:22.12 |\n| 2008 | World Indoor Championships | Valencia, Spain | 5th | 1500 m | 4:15.54 |\n| 2008 | Olympic Games | Beijing, China | 10th | 1500 m | 4:05.57 |\n| 2009 | Mediterranean Games | Pescara, Italy | 4th | 1500 m | 4:12.83 |\n| 2009 | World Championships | Berlin, Germany | 29th (h) | 1500 m | 4:10.57 |\n| 2009 | Jeux de la Francophonie | Beirut, Lebanon | 2nd | 1500 m | 4:21.56 |\n| 2010 | African Championships | Nairobi, Kenya | 11th | 1500 m | 4:20.98 |\n| 2011 | World Championships | Daegu, South Korea | 16th (sf) | 1500 m | 4:09.64 |\n| 2011 | Pan Arab Games | Doha, Qatar | 2nd | 1500 m | 4:20.83 |\n| 2012 | World Indoor Championships | Istanbul, Turkey | 11th (h) | 1500 m | 4:11.69 |\n| 2013 | Mediterranean Games | Mersin, Turkey | 2nd | 800 m | 2:00.79 |\n| 2013 | Mediterranean Games | Mersin, Turkey | 1st | 1500 m | 4:04.06 |\n| 2013 | World Championships | Moscow, Russia | 11th | 1500 m | 4:09.16 |\n| 2013 | Jeux de la Francophonie | Nice, France | 4th | 800 m | 2:03.73 |\n| 2013 | Jeux de la Francophonie | Nice, France | 2nd | 1500 m | 4:18.89 |\n| 2014 | World Indoor Championships | Sopot, Poland | 4th | 1500 m | 4:07.62 |\n\n\nQuestion\nhow many races did she compete in each of the following years: 2009, 2010, and 2011.\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"6"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-6141",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-16.csv` in your code execution environment.\n\nTable Title: 1982 Winnipeg Blue Bombers season\nTable Content:\n| Week | Date | Opponent | Score | Result | Record |\n| --- | --- | --- | --- | --- | --- |\n| 1 | July 9 | vs. Saskatchewan Roughriders | 31–21 | Win | 1–0 |\n| 2 | July 16 | at Montreal Concordes | 36–0 | Win | 2–0 |\n| 3 | July 24 | vs. Hamilton Tiger-Cats | 36–25 | Loss | 2–1 |\n| 4 | Aug 1 | at Edmonton Eskimos | 32–26 | Win | 3–1 |\n| 5 | Aug 8 | vs. BC Lions | 29–16 | Win | 4–1 |\n| 6 | Aug 20 | at Calgary Stampeders | 35–4 | Win | 5–1 |\n| 7 | Aug 28 | vs. Ottawa Rough Riders | 27–20 | Win | 6–1 |\n| 8 | Sept 5 | at Saskatchewan Roughriders | 36–35 | Win | 7–1 |\n| 9 | Sept 11 | vs. Calgary Stampeders | 15–11 | Loss | 7–2 |\n| 10 | Sept 18 | at Ottawa Rough Riders | 38–28 | Loss | 7–3 |\n| 11 | Sept 25 | vs. Montreal Concordes | 19–16 | Win | 8–3 |\n| 12 | Oct 2 | at BC Lions | 29–19 | Win | 9–3 |\n| 13 | Oct 11 | vs. Toronto Argonauts | 39–35 | Win | 10–3 |\n| 14 | Oct 17 | at Toronto Argonauts | 29–16 | Win | 11–3 |\n| 15 | Oct 23 | vs. Edmonton Eskimos | 33–17 | Loss | 11–4 |\n| 16 | Nov 7 | at Hamilton Tiger-Cats | 24–21 | Loss | 11–5 |\n\n\nQuestion\nhow many total points did the bombers score against the bc lions?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"58"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-16",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-7551.csv` in your code execution environment.\n\nTable Title: Human rights in Niger\nTable Content:\n| Year | Political Rights | Civil Liberties | Status | President |\n| --- | --- | --- | --- | --- |\n| 1972 | 6 | 6 | Not Free | Hamani Diori |\n| 1973 | 6 | 6 | Not Free | Hamani Diori |\n| 1974 | 7 | 6 | Not Free | Hamani Diori |\n| 1975 | 7 | 6 | Not Free | Seyni Kountché |\n| 1976 | 7 | 6 | Not Free | Seyni Kountché |\n| 1977 | 7 | 6 | Not Free | Seyni Kountché |\n| 1978 | 7 | 6 | Not Free | Seyni Kountché |\n| 1979 | 7 | 6 | Not Free | Seyni Kountché |\n| 1980 | 7 | 6 | Not Free | Seyni Kountché |\n| 1981 | 7 | 6 | Not Free | Seyni Kountché |\n| 1982 | 7 | 6 | Not Free | Seyni Kountché |\n| 1983 | 7 | 6 | Not Free | Seyni Kountché |\n| 1984 | 7 | 6 | Not Free | Seyni Kountché |\n| 1985 | 7 | 6 | Not Free | Seyni Kountché |\n| 1986 | 7 | 6 | Not Free | Seyni Kountché |\n| 1987 | 7 | 6 | Not Free | Seyni Kountché |\n| 1988 | 6 | 6 | Not Free | Ali Saibou |\n| 1989 | 7 | 6 | Not Free | Ali Saibou |\n| 1990 | 6 | 5 | Not Free | Ali Saibou |\n| 1991 | 6 | 5 | Partly Free | Ali Saibou |\n| 1992 | 5 | 4 | Partly Free | Ali Saibou |\n| 1993 | 3 | 4 | Partly Free | Ali Saibou |\n| 1994 | 3 | 5 | Partly Free | Mahamane Ousmane |\n| 1995 | 3 | 5 | Partly Free | Mahamane Ousmane |\n| 1996 | 7 | 5 | Not Free | Mahamane Ousmane |\n| 1997 | 7 | 5 | Not Free | Ibrahim Baré Maïnassara |\n| 1998 | 7 | 5 | Not Free | Ibrahim Baré Maïnassara |\n| 1999 | 5 | 5 | Partly Free | Ibrahim Baré Maïnassara |\n| 2000 | 4 | 4 | Partly Free | Mamadou Tandja |\n| 2001 | 4 | 4 | Partly Free | Mamadou Tandja |\n| 2002 | 4 | 4 | Partly Free | Mamadou Tandja |\n| 2003 | 4 | 4 | Partly Free | Mamadou Tandja |\n| 2004 | 3 | 3 | Partly Free | Mamadou Tandja |\n| 2005 | 3 | 3 | Partly Free | Mamadou Tandja |\n| 2006 | 3 | 3 | Partly Free | Mamadou Tandja |\n| 2007 | 3 | 4 | Partly Free | Mamadou Tandja |\n| 2008 | 3 | 4 | Partly Free | Mamadou Tandja |\n| 2009 | 5 | 4 | Partly Free | Mamadou Tandja |\n| 2010 | 5 | 4 | Partly Free | Mamadou Tandja |\n| 2011 | 3 | 4 | Partly Free | Salou Djibo |\n\n\nQuestion\nwho ruled longer, ali saibou or mamadou tandja?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"Mamadou Tandja"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-7551",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-5985.csv` in your code execution environment.\n\nTable Title: 2009–10 Fresno State Bulldogs men's basketball team\nTable Content:\n| # | Name | Height | Weight (lbs.) | Position | Class | Hometown | Previous Team(s) |\n| --- | --- | --- | --- | --- | --- | --- | --- |\n| 1 | Mychal Ladd | 6'5\" | 190 | G | So. | Seattle, WA, U.S. | Rainier Beach HS |\n| 2 | Jonathan Wills | 6'6\" | 180 | G | Jr. | Carson, CA, U.S. | Mayfair HS\\nNew Mexico |\n| 3 | Taylor Kelly | 6'2\" | 220 | G | Fr. | Bakersfield, CA, U.S. | Centennial HS |\n| 11 | Jerry Brown, Jr. | 6'7\" | 195 | G\\nF | Fr. | Richmond, CA, U.S. | Sacred Heart |\n| 12 | Steven Shepp | 6'2\" | 170 | G | So. | Woodbridge, Va, U.S. | C.D. Hylton HS\\nAllegany College of Maryland |\n| 15 | Nedeljko Golubovic | 6'8\" | 245 | F | Jr. | Pljevlja, Serbia, Serbia | Findlay Prep (Las Vegas, NV) |\n| 20 | Garrett Johnson | 6'5\" | 190 | G\\nF | Fr. | Pomona, CA, U.S. | Diamond Ranch HS |\n| 22 | Greg Smith | 6'10\" | 250 | C | Fr. | Fresno, CA, U.S. | Westwind Academy |\n| 24 | Paul George | 6'8\" | 210 | G\\nF | So. | Palmdale, CA, U.S. | Pete Knight HS |\n| 25 | Justin Johnson | 6'6\" | 205 | F | Fr. | Phoenix, AZ, U.S. | Brophy College Prep |\n| 30 | Sylvester Seay | 6'10\" | 235 | F | Sr. | San Bernardino, CA, U.S. | Winchendon School\\nArizona State |\n| 32 | Brandon Sperling | 6'4\" | 180 | G | So. | Fresno, CA, U.S. | Buchanan HS |\n| 42 | Brandon Webster | 6'10\" | 275 | C | Jr. | Bryan, TX, U.S. | Bryan HS |\n\n\nQuestion\nwho weighs the most on the team?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"Brandon Webster"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-5985",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-3175.csv` in your code execution environment.\n\nTable Title: Sailing at the 2008 Summer Olympics – Qualification\nTable Content:\n| # | Nation | Qualification Tournament | Place in QT | Sailor |\n| --- | --- | --- | --- | --- |\n| 1 | China | Host Country | | |\n| 2 | Brazil | 2007 Worlds | 1 | Robert Scheidt, Bruno Prada |\n| 3 | France | 2007 Worlds | 2 | Xavier Rohart, Pascal Rambeau |\n| 4 | Great Britain | 2007 Worlds | 3 | Iain Percy, Andrew Simpson |\n| 5 | New Zealand | 2007 Worlds | 4 | Hamish Pepper, Carl Williams |\n| 6 | Italy | 2007 Worlds | 5 | Diego Negri, Luigi Viale |\n| 7 | Poland | 2007 Worlds | 6 | Mateusz Kusznierewicz, Dominik Zycki |\n| 8 | Germany | 2007 Worlds | 7 | Marc Pickel, Ingo Borkowski |\n| 9 | Sweden | 2007 Worlds | 8 | Fredrik Lööf, Anders Ekström |\n| 10 | Australia | 2007 Worlds | 9 | Iain Murray, Andrew Palfrey |\n| 11 | Portugal | 2007 Worlds | 10 | Afonso Domingos, Bernardo Santos |\n| 12 | United States | 2007 Worlds | 11 | Mark Reynolds, Hal Haenel |\n| 13 | Switzerland | 2008 Worlds | 4 | Flavio Marazzi, Enrico De Maria |\n| 14 | Croatia | 2008 Worlds | 12 | Marin Lovrovic, Sinsa Mikulicic |\n| 15 | Ireland | 2008 Worlds | 14 | Max Treacy, Anthony Shanks |\n| 16 | Austria | 2008 Worlds | 16 | Hans Spitzauer, Christian Nehammer |\n\n\nQuestion\nwhich nation had their sailors as robert scheidt and bruno prada?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"Brazil"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-3175",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-3201.csv` in your code execution environment.\n\nTable Title: Michael W. Smith discography\nTable Content:\n| Year | Album | Peak\\nUS | Peak\\nUS\\nHoliday | Certifications\\n(sales threshold) |\n| --- | --- | --- | --- | --- |\n| 1990 | Go West Young Man\\n\\nReleased: October 1, 1990\\nLabel: Reunion\\nFormat: CD | 74 | — | RIAA: Platinum\\nMC: Gold |\n| 1992 | Change Your World\\n\\nReleased: September 1, 1992\\nLabel: Reunion\\nFormat: CD | 86 | — | RIAA: Platinum |\n| 1995 | I'll Lead You Home\\n\\nReleased: August 29, 1995\\nLabel: Reunion\\nFormat: CD | 16 | — | RIAA: Platinum |\n| 1998 | Live the Life\\n\\nReleased: April 28, 1998\\nLabel: Reunion\\nFormat: CD | 23 | — | RIAA: Gold |\n| 1998 | Christmastime\\n\\nReleased: October 13, 1998\\nLabel: Reunion\\nFormat: CD | 90 | 4 | RIAA: Gold |\n| 1999 | This Is Your Time\\n\\nReleased: November 23, 1999\\nLabel: Reunion\\nFormat: CD | 21 | — | RIAA: Gold |\n\n\nQuestion\nwhat michael w smith album was released before his christmastime album?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"Live the Life"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-3201",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-14133.csv` in your code execution environment.\n\nTable Title: European Baseball Championship\nTable Content:\n| Rank | Nation | Gold | Silver | Bronze | Total |\n| --- | --- | --- | --- | --- | --- |\n| 1 | Netherlands | 20 | 9 | 0 | 29 |\n| 2 | Italy | 10 | 15 | 3 | 28 |\n| 3 | Belgium | 1 | 2 | 6 | 9 |\n| 4 | Spain | 1 | 1 | 13 | 15 |\n| 5 | Great Britain | 0 | 2 | 0 | 2 |\n| 6 | Germany | 0 | 1 | 7 | 8 |\n| 7 | Greece | 0 | 1 | 0 | 1 |\n| 7 | Russia | 0 | 1 | 0 | 1 |\n| 9 | Sweden | 0 | 0 | 2 | 2 |\n| 10 | France | 0 | 0 | 1 | 1 |\n\n\nQuestion\nonly nation to earn more than 7 bronze medals\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"Spain"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-14133",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-4017.csv` in your code execution environment.\n\nTable Title: Niobium\nTable Content:\n| Country | 2000 | 2001 | 2002 | 2003 | 2004 | 2005 | 2006 | 2007 | 2008 | 2009 | 2010 | 2011 |\n| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |\n| Australia | 160 | 230 | 290 | 230 | 200 | 200 | 200 | ? | ? | ? | ? | ? |\n| Brazil | 30,000 | 22,000 | 26,000 | 29,000 | 29,900 | 35,000 | 40,000 | 57,300 | 58,000 | 58,000 | 58,000 | 58,000 |\n| Canada | 2,290 | 3,200 | 3,410 | 3,280 | 3,400 | 3,310 | 4,167 | 3,020 | 4,380 | 4,330 | 4,420 | 4,400 |\n| Congo D.R. | ? | 50 | 50 | 13 | 52 | 25 | ? | ? | ? | ? | ? | ? |\n| Mozambique | ? | ? | 5 | 34 | 130 | 34 | 29 | ? | ? | ? | ? | ? |\n| Nigeria | 35 | 30 | 30 | 190 | 170 | 40 | 35 | ? | ? | ? | ? | ? |\n| Rwanda | 28 | 120 | 76 | 22 | 63 | 63 | 80 | ? | ? | ? | ? | ? |\n| World | 32,600 | 25,600 | 29,900 | 32,800 | 34,000 | 38,700 | 44,500 | 60,400 | 62,900 | 62,900 | 62,900 | 63,000 |\n\n\nQuestion\nafter brazil which country produced the most niobium?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"Canada"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-4017",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-11225.csv` in your code execution environment.\n\nTable Title: List of highest-grossing openings for films\nTable Content:\n| Rank | Film | Year | Opening Weekend\\n(three day) | Inflation-adjusted\\n(2014 USD) |\n| --- | --- | --- | --- | --- |\n| 1 | Marvel's The Avengers | 2012 | $207,438,708 | $213,091,480 |\n| 2 | Iron Man 3 | 2013 | $174,144,585 | $176,308,953 |\n| 3 | Harry Potter and the Deathly Hallows – Part 2 | 2011 | $169,189,427 | $177,373,590 |\n| 4 | The Dark Knight Rises | 2012 | $160,887,295 | $165,271,526 |\n| 5 | The Dark Knight | 2008 | $158,411,483 | $173,518,111 |\n| 6 | The Hunger Games: Catching Fire | 2013 | $158,074,286 | $160,038,924 |\n| 7 | The Hunger Games | 2012 | $152,535,747 | $156,692,396 |\n| 8 | Spider-Man 3 | 2007 | $151,116,516 | $171,876,545 |\n| 9 | The Twilight Saga: New Moon | 2009 | $142,839,137 | $157,018,918 |\n| 10 | The Twilight Saga: Breaking Dawn - Part 2 | 2012 | $141,067,634 | $144,911,773 |\n| 11 | The Twilight Saga: Breaking Dawn - Part 1 | 2011 | $138,122,261 | $144,803,619 |\n| 12 | Pirates of the Caribbean: Dead Man's Chest | 2006 | $135,634,554 | $158,673,173 |\n| 13 | Iron Man 2 | 2010 | $128,122,480 | $138,563,103 |\n| 14 | Harry Potter and the Deathly Hallows – Part 1 | 2010 | $125,017,372 | $135,204,962 |\n| 15 | Shrek the Third | 2007 | $121,629,270 | $138,338,411 |\n| 16 | Man of Steel | 2013 | $116,619,362 | $118,068,774 |\n| 17 | Alice in Wonderland | 2010 | $116,101,023 | $125,562,025 |\n| 18 | Spider-Man | 2002 | $114,844,116 | $150,582,840 |\n| 19 | Pirates of the Caribbean: At World's End | 2007 | $114,732,820 | $130,494,543 |\n| 20 | Toy Story 3 | 2010 | $110,307,189 | $119,296,055 |\n| 21 | Transformers: Revenge of the Fallen | 2009 | $108,966,307 | $119,783,499 |\n| 22 | Star Wars Episode III: Revenge of the Sith | 2005 | $108,435,841 | $130,939,650 |\n| 23 | Shrek 2 | 2004 | $108,037,878 | $134,895,074 |\n| 24 | X-Men: The Last Stand | 2006 | $102,750,665 | $120,203,691 |\n| 25 | Harry Potter and the Goblet of Fire | 2005 | $102,685,961 | $123,996,491 |\n| 26 | Indiana Jones and the Kingdom of the Crystal Skull | 2008 | $100,137,835 | $109,687,301 |\n\n\nQuestion\nname one that ranked after man of steel.\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"Alice in Wonderland"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-11225",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-13686.csv` in your code execution environment.\n\nTable Title: 2007 Houston Astros season\nTable Content:\n| # | Date | Opponent | Score | Win | Loss | Save | Attendance | Record |\n| --- | --- | --- | --- | --- | --- | --- | --- | --- |\n| 136 | September 1 | @ Cubs | 4 - 3 | Marquis (11-8) | Patton (0-2) | Dempster (24) | 40,606 | 61-75 |\n| 137 | September 2 | @ Cubs | 6 - 5 | Marmol (5-1) | Qualls (6-4) | Dempster (25) | 41,415 | 61-76 |\n| 138 | September 3 | @ Brewers | 9 - 7 | Borkowski (4-3) | Aquino (0-1) | Qualls (4) | 31,226 | 62-76 |\n| 139 | September 4 | @ Brewers | 5 - 3 | Villanueva (7-3) | Backe (0-1) | Cordero (40) | 25,854 | 62-77 |\n| 140 | September 5 | @ Brewers | 14 - 2 | Gallardo (7-4) | Albers (4-7) | | 28,988 | 62-78 |\n| 141 | September 7 | @ Mets | 11 - 3 | Pelfrey (2-7) | Rodríguez (8-13) | | 51,113 | 62-79 |\n| 142 | September 8 | @ Mets | 3 - 1 | Glavine (13-6) | Williams (8-14) | Wagner (31) | 53,061 | 62-80 |\n| 143 | September 9 | @ Mets | 4 - 1 | Martínez (2-0) | Oswalt (14-7) | Wagner (32) | 51,847 | 62-81 |\n| 144 | September 11 | Cubs | 5 - 4 (11) | Lidge (4-2) | Dempster (2-6) | | 33,493 | 63-81 |\n| 145 | September 12 | Cubs | 3 - 2 | Hill (9-8) | Albers (4-8) | Dempster (26) | 33,115 | 63-82 |\n| 146 | September 13 | Cubs | 6 - 2 | Trachsel (7-10) | Williams (8-15) | | 34,234 | 63-83 |\n| 147 | September 14 | Pirates | 4 - 3 | Sánchez (1-0) | Qualls (6-5) | Capps (17) | 35,352 | 63-84 |\n| 148 | September 15 | Pirates | 9 - 7 | Gutiérrez (1-1) | Morris (9-10) | Lidge (15) | 40,425 | 64-84 |\n| 149 | September 16 | Pirates | 15 - 3 | Backe (1-1) | Maholm (10-15) | | 35,715 | 65-84 |\n| 150 | September 17 | Brewers | 6 - 0 | Gallardo (9-4) | Albers (4-9) | | 32,578 | 65-85 |\n| 151 | September 18 | Brewers | 9 - 1 | Vargas (11-4) | Paulino (0-1) | | 32,866 | 65-86 |\n| 152 | September 19 | Brewers | 5 - 4 (10) | Lidge (5-2) | Wise (3-2) | | 36,981 | 66-86 |\n| 153 | September 20 | @ Cardinals | 18 - 1 | Rodríguez (9-13) | Looper (12-11) | | 42,171 | 67-86 |\n| 154 | September 21 | @ Cardinals | 6 - 3 | Backe (2-1) | Piñeiro (6-5) | Lidge (16) | 43,677 | 68-86 |\n| 155 | September 22 | @ Cardinals | 7 - 4 | Thompson (7-6) | Albers (4-10) | Isringhausen (30) | 46,237 | 68-87 |\n| 156 | September 23 | @ Cardinals | 4 - 3 | Jiménez (3-0) | Lidge (5-3) | | 46,169 | 68-88 |\n| 157 | September 25 | @ Reds | 8 - 5 | Paulino (1-1) | Coutlangus (4-2) | Lidge (17) | 13,261 | 69-88 |\n| 158 | September 26 | @ Reds | 7 - 6 | Sarfate (1-0) | Burton (4-2) | Qualls (5) | 13,138 | 70-88 |\n| 159 | September 27 | @ Reds | 4 - 3 | Borkowski (5-3) | Majewski (0-4) | Lidge (18) | 13,626 | 71-88 |\n| 160 | September 28 | Braves | 7 - 2 | Reyes (2-2) | Albers (4-11) | | 43,011 | 71-89 |\n| 161 | September 29 | Braves | 3 - 2 | Backe (3-1) | Bennett (2-1) | Lidge (19) | 43,624 | 72-89 |\n| 162 | September 30 | Braves | 3 - 0 | Paulino (2-1) | Carlyle (8-7) | Borkowski (1) | 43,823 | 73-89 |\n\n\nQuestion\nwho did the astros play after the reds?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"Braves"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-13686",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-7475.csv` in your code execution environment.\n\nTable Title: New York gubernatorial election, 2014\nTable Content:\n| Poll source | Date(s)\\nadministered | Sample\\nsize | Margin of\\nerror | Andrew\\nCuomo (D) | Rob\\nAstorino (R) | Other | Undecided |\n| --- | --- | --- | --- | --- | --- | --- | --- |\n| Siena | March 16–20, 2014 | 813 | ± 3.4% | 61% | 26% | — | 13% |\n| Marist | February 28–March 3, 2014 | 658 | ± 3.8% | 65% | 25% | — | 10% |\n| Quinnipiac | February 6–10, 2014 | 1,488 | ± 2.5% | 58% | 24% | 2% | 16% |\n| Siena | January 12–16, 2014 | 808 | ± 3.4% | 67% | 19% | 3% | 11% |\n| Quinnipiac | November 20–24, 2013 | 1,337 | ± 2.7% | 56% | 25% | 2% | 17% |\n| Marist | November 18–20, 2013 | 675 | ± 3.8% | 65% | 23% | — | 12% |\n| Siena | November 11–14, 2013 | 806 | ± 3.5% | 63% | 24% | — | 13% |\n\n\nQuestion\nwhat was the difference in percentage between cuomo and astorino from february 28-march 3, 2014?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"40%"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-7475",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-11304.csv` in your code execution environment.\n\nTable Title: Nagoya Municipal Subway\nTable Content:\n| Color & Icon | Name | Name | Mark | First section\\nopened | Last ex-\\ntension | Length | Stations |\n| --- | --- | --- | --- | --- | --- | --- | --- |\n| yellow | Line 1 | Higashiyama Line | H | 1957 | 1982 | 20.6 km (12.8 mi) | 22 |\n| purple | Line 2 | Meijō Line | M | 1965 | 1971 | 8.9 km (5.5 mi) | 12 |\n| purple | Line 4 | Meijō Line | M | 1974 | 2004 | 17.5 km (10.9 mi) | 17 |\n| purple & white | Line 2 | Meikō Line | E | 1971 | - | 6.0 km (3.7 mi) | 7 |\n| blue | Via trackage rights | Meitetsu Inuyama Line | - | 1993 | - | 21.4 km (13.3 mi) | 13 |\n| blue | Line 3 | Tsurumai Line | T | 1977 | 1993 | 20.4 km (12.7 mi) | 20 |\n| blue | Via trackage rights | Meitetsu Toyota Line | - | 1979 | - | 15.2 km (9.4 mi) | 8 |\n| blue | Via trackage rights | Meitetsu Mikawa Line | - | N/A | - | 1.4 km (0.87 mi) | 2 |\n| red | Line 6 | Sakura-dōri Line | S | 1989 | 2011 | 19.1 km (11.9 mi) | 21 |\n| pink | Via trackage rights | Meitetsu Komaki Line | - | 2003 | - | 18.3 km (11.4 mi) | 13 |\n| pink | Via trackage rights | Meitetsu Komaki Line | - | 2003 | - | 2.3 km (1.4 mi) | 2 |\n| pink | Line 7 | Kamiiida Line | K | 2003 | - | 0.8 km (0.50 mi) | 2 |\n\n\nQuestion\nis the higashiyama line longer than the meiko line?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"Yes"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-11304",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-8058.csv` in your code execution environment.\n\nTable Title: National Democratic Party of Germany\nTable Content:\n| Election year | # of\\nconstituency votes | % of\\nconstituency votes | +/– | # of\\nparty list votes | % of\\nparty list votes | +/– | # of\\noverall seats won |\n| --- | --- | --- | --- | --- | --- | --- | --- |\n| 1965 | 587,216 | 1.8 | 1.8 | 664,193 | 2.0 | 2.0 | 0 / 518 |\n| 1969 | 1,189,375 | 3.6 | 1.8 | 1,422,010 | 4.3 | 2.3 | 0 / 518 |\n| 1972 | 194,389 | 0.5 | 3.1 | 207,465 | 0.6 | 3.7 | 0 / 518 |\n| 1976 | 136.023 | 0.4 | 0.1 | 122,661 | 0.3 | 0.3 | 0 / 518 |\n| 1980 | | | | 68,096 | 0.2 | 0.1 | 0 / 497 |\n| 1983 | 57,112 | 0.1 | 0.3 | 91,095 | 0.2 | 0 | 0 / 498 |\n| 1987 | 182,880 | 0.5 | 0.4 | 227,054 | 0.6 | 0.4 | 0 / 497 |\n| 1990 | 190,105 | 0.4 | 0.1 | 145,776 | 0.3 | 0.3 | 0 / 662 |\n| 1998 | 45,043 | 0.1 | 0.3 | 126,571 | 0.3 | 0 | 0 / 669 |\n| 2002 | 103,209 | 0.1 | 0.1 | 215,232 | 0.4 | 0.1 | 0 / 603 |\n| 2005 | 857.777 | 1.8 | 1.6 | 748,568 | 1.6 | 1.2 | 0 / 614 |\n| 2009 | 768,442 | 1.8 | 0 | 635,525 | 1.5 | 0.1 | 0 / 620 |\n| 2013 | 634,842 | 1.5 | 0.3 | 560,660 | 1.3 | 0.2 | 0 / 630 |\n\n\nQuestion\nafter 1983 when is the next time the % of constituency was 0.1?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"1998"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-8058",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-1565.csv` in your code execution environment.\n\nTable Title: FireHouse (band)\nTable Content:\n| Year | Song | US Hot 100 | US MSR | US A.C. | CAN | UK Singles | Album |\n| --- | --- | --- | --- | --- | --- | --- | --- |\n| 1991 | \"Don't Treat Me Bad\" | 19 | 16 | - | 35 | 71 | FireHouse |\n| 1991 | \"Love of a Lifetime\" | 5 | - | - | 30 | 40 | FireHouse |\n| 1991 | \"All She Wrote\" | 58 | 25 | - | - | 56 | FireHouse |\n| 1992 | \"Reach for the Sky\" | 83 | 27 | - | - | 67 | Hold Your Fire |\n| 1992 | \"When I Look Into Your Eyes\" | 8 | - | - | 20 | 65 | Hold Your Fire |\n| 1992 | \"Sleeping with You\" | 78 | - | - | - | 40 | Hold Your Fire |\n| 1995 | \"I Live My Life for You\" | 26 | - | 20 | 17 | 43 | 3 |\n| 1995 | \"Here For You\" | 108 | - | - | - | - | 3 |\n\n\nQuestion\nwhat year did firehouse release their last single?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"1995"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-1565",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-2157.csv` in your code execution environment.\n\nTable Title: Electricity in Sri Lanka\nTable Content:\n| | 2000 | 2001 | 2002 | 2003 | 2004 | 2005 | 2006 | 2007 | 2008 | 2009 | 2010 | 2011 | 2012 |\n| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |\n| Hydro power | 1,150 | 1,161 | 1,171 | 1,247 | 1,281 | 1,293 | 1,316 | 1,326 | 1,357 | 1,379 | 1,382 | 1,401 | 1,584 |\n| Thermal | 685 | 835 | 1,056 | 1,233 | 1,215 | 1,155 | 1,155 | 1,155 | 1,285 | 1,290 | 1,390 | 1,690 | 1,638 |\n| Other renewables | 3 | 3 | 3 | 3 | 3 | 3 | 3 | 3 | 3 | 15 | 45 | 50 | 90 |\n| Total availability | 1,838 | 1,999 | 2,230 | 2,483 | 2,499 | 2,411 | 2,434 | 2,444 | 2,645 | 2,684 | 2,818 | 3,141 | 3,312 |\n| Hydro power | 3,197 | 3,110 | 2,962 | 3,310 | 2,960 | 3,451 | 4,634 | 3,947 | 4,130 | 3,881 | 5,634 | 4,622 | 3,292 |\n| Thermal | 3,485 | 3,407 | 3,201 | 4,298 | 5,080 | 5,314 | 4,751 | 5,864 | 5,763 | 5,975 | 4,995 | 6,785 | 8,339 |\n| Other renewables | 3 | 3 | 4 | 3 | 3 | 5 | 4 | 4 | 8 | 27 | 86 | 121 | 169 |\n| Total production | 6,685 | 6,520 | 6,167 | 7,611 | 8,043 | 8,770 | 9,389 | 9,815 | 9,901 | 9,883 | 10,715 | 11,528 | 11,800 |\n\n\nQuestion\nwhat kind of power has the least amount of capacity in 2007?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"Other renewables"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-2157",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-6715.csv` in your code execution environment.\n\nTable Title: 1979–80 NBA season\nTable Content:\n| Central Division\\n\\n\\nv\\nt\\ne | W | L | PCT | GB | Home | Road | Div |\n| --- | --- | --- | --- | --- | --- | --- | --- |\n| y-Atlanta Hawks | 50 | 32 | .610 | – | 32–9 | 18–23 | 21–9 |\n| x-Houston Rockets | 41 | 41 | .500 | 9 | 29–12 | 12–29 | 20–10 |\n| x-San Antonio Spurs | 41 | 41 | .500 | 9 | 27–14 | 14–27 | 14–16 |\n| Cleveland Cavaliers | 37 | 45 | .451 | 13 | 28–13 | 9–32 | 16–14 |\n| Indiana Pacers | 37 | 45 | .451 | 13 | 26–15 | 11–30 | 15–15 |\n| Detroit Pistons | 16 | 66 | .195 | 34 | 13–28 | 3–38 | 4–26 |\n\n\nQuestion\nhow many teams have the same amount of losses (l)?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"4"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-6715",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-5164.csv` in your code execution environment.\n\nTable Title: Playoff Bowl\nTable Content:\n| Season | Network | Play-by-play | Color commentator(s) | Sideline reporter(s) |\n| --- | --- | --- | --- | --- |\n| 1969 | CBS | Jack Whitaker | Frank Gifford and Don Perkins | |\n| 1968 | CBS | Ray Scott | Paul Christman | Frank Glieber |\n| 1967 | CBS | Frank Glieber | Frank Gifford | |\n| 1966 | CBS | Chuck Thompson | Tom Brookshier | |\n| 1965 | CBS | Frank Glieber (first half) and Chuck Thompson (second half) | Pat Summerall | |\n| 1964 | CBS | Jack Drees (first half) and Earl Gillespie (second half) | Frank Gifford | |\n| 1963 | CBS | Ray Scott (first half) and Ken Coleman (second half) | Frank Gifford | |\n| 1962 | CBS | Chris Schenkel (first half) and Ray Scott (second half) | Warren Lahr | |\n| 1961 | CBS | Chris Schenkel (first half) and Van Patrick (second half) | Johnny Lujack | |\n\n\nQuestion\nwho was the last color commentator?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"Frank Gifford and Don Perkins"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-5164",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-1742.csv` in your code execution environment.\n\nTable Title: Ministry of Development (Greece)\nTable Content:\n| | Name | Took Office | Left Office | Party |\n| --- | --- | --- | --- | --- |\n| 1 | Vasso Papandreou | 21 February 1996 | 19 February 1999 | Panhellenic Socialist Movement |\n| 2 | Evangelos Venizelos | 19 February 1999 | 13 April 2000 | Panhellenic Socialist Movement |\n| 3 | Nikos Christodoulakis | 13 April 2000 | 24 October 2001 | Panhellenic Socialist Movement |\n| 4 | Akis Tsochatzopoulos | 24 October 2001 | 10 March 2004 | Panhellenic Socialist Movement |\n| 5 | Dimitris Sioufas | 10 March 2004 | 19 September 2007 | New Democracy |\n| 6 | Christos Folias | 19 September 2007 | 8 January 2009 | New Democracy |\n| 7 | Kostis Hatzidakis | 8 January 2009 | 7 October 2009 | New Democracy |\n\n\nQuestion\nwho took office in 2000?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"Nikos Christodoulakis"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-1742",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-872.csv` in your code execution environment.\n\nTable Title: Jalkapalloseura Airport City\nTable Content:\n| Season | Level | Division | Section | Administration | Position | Movements |\n| --- | --- | --- | --- | --- | --- | --- |\n| 1999 | Tier 2 | Ykkönen (First Division) | South Group | Finnish FA (Suomen Pallolitto) | 9th | Relegation Group - Relegated |\n| 2000 | Tier 3 | Kakkonen (Second Division) | South Group | Finnish FA (Suomen Pallolitto) | 2nd | Play-Offs - Promoted |\n| 2001 | Tier 2 | Ykkönen (First Division) | South Group | Finnish FA (Suomen Pallolitto) | 9th | Relegated |\n| 2002 | Tier 3 | Kakkonen (Second Division) | South Group | Finnish FA (Suomen Pallolitto) | 2nd | |\n| 2003 | Tier 3 | Kakkonen (Second Division) | South Group | Finnish FA (Suomen Pallolitto) | 11th | Relegated |\n| 2004 | Tier 4 | Kolmonen (Third Division) | Section 3 | Helsinki & Uusimaa (SPL Helsinki) | 8th | |\n| 2005 | Tier 4 | Kolmonen (Third Division) | Section 3 | Helsinki & Uusimaa (SPL Helsinki) | 6th | |\n| 2006 | Tier 4 | Kolmonen (Third Division) | Section 2 | Helsinki & Uusimaa (SPL Uusimaa) | 7th | |\n| 2007 | Tier 4 | Kolmonen (Third Division) | Section 3 | Helsinki & Uusimaa (SPL Uusimaa) | 2nd | |\n| 2008 | Tier 4 | Kolmonen (Third Division) | Section 3 | Helsinki & Uusimaa (SPL Helsinki) | 2nd | |\n| 2009 | Tier 4 | Kolmonen (Third Division) | Section 2 | Helsinki & Uusimaa (SPL Helsinki) | 1st | Promoted |\n| 2010 | Tier 3 | Kakkonen (Second Division) | Group A | Finnish FA (Suomen Pallolitto) | | |\n\n\nQuestion\nhow many 2nd positions were there?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"4"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-872",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-7357.csv` in your code execution environment.\n\nTable Title: Israeli Basketball League Cup\nTable Content:\n| Place | Team | Titles | Runners-up | Winning Years |\n| --- | --- | --- | --- | --- |\n| 1 | Maccabi Tel Aviv | 5 | 1 | 2007, 2010, 2011, 2012, 2013 |\n| 2 | Hapoel Jerusalem | 2 | 3 | 2008, 2009 |\n| 3 | Ironi Ashkelon | 1 | 0 | 2006 |\n| 4 | Maccabi Rishon LeZion | 0 | 1 | |\n| 4 | Ironi Nahariya | 0 | 1 | |\n| 4 | Hapoel Holon | 0 | 1 | |\n| 4 | Maccabi Ashdod | 0 | 1 | |\n\n\nQuestion\nwhich team has the most titles?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"Maccabi Tel Aviv"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-7357",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-6708.csv` in your code execution environment.\n\nTable Title: 1996 Indianapolis Colts season\nTable Content:\n| Week | Date | Opponent | Result | Record | Game Site | TV Time | Attendance |\n| --- | --- | --- | --- | --- | --- | --- | --- |\n| 1 | September 1, 1996 | Arizona Cardinals | W 20–13 | 1–0 | RCA Dome | FOX 1:00 pm | 48,133 |\n| 2 | September 8, 1996 | at New York Jets | W 21–7 | 2–0 | The Meadowlands | NBC 4:15 pm | 63,534 |\n| 3 | September 15, 1996 | at Dallas Cowboys | W 25–24 | 3–0 | Texas Stadium | NBC 4:15 pm | 63,021 |\n| 4 | September 23, 1996 | Miami Dolphins | W 10–6 | 4–0 | RCA Dome | ABC 9:00 pm | 60,891 |\n| 5 | Bye | Bye | Bye | Bye | Bye | Bye | Bye |\n| 6 | October 6, 1996 | at Buffalo Bills | L 13–16 | 4–1 | Rich Stadium | NBC 4:15 pm | 79,401 |\n| 7 | October 13, 1996 | Baltimore Ravens | W 26–21 | 5–1 | RCA Dome | TNT 8:15 pm | 56,978 |\n| 8 | October 20, 1996 | New England Patriots | L 9–27 | 5–2 | RCA Dome | NBC 1:00 pm | 58,725 |\n| 9 | October 27, 1996 | at Washington Redskins | L 16–31 | 5–3 | RFK Stadium | NBC 1:00 pm | 54,254 |\n| 10 | November 3, 1996 | San Diego Chargers | L 19–26 | 5–4 | RCA Dome | NBC 1:00 pm | 58,484 |\n| 11 | November 10, 1996 | at Miami Dolphins | L 13–37 | 5–5 | Joe Robbie Stadium | NBC 1:00 pm | 66,623 |\n| 12 | November 17, 1996 | New York Jets | W 34–29 | 6–5 | RCA Dome | NBC 1:00 pm | 48,322 |\n| 13 | November 24, 1996 | at New England Patriots | L 13–27 | 6–6 | Foxboro Stadium | NBC 1:00 pm | 58,226 |\n| 14 | December 1, 1996 | Buffalo Bills | W 13–10 | 7–6 | RCA Dome | NBC 4:15 pm | 53,804 |\n| 15 | December 5, 1996 | Philadelphia Eagles | W 37–10 | 8–6 | RCA Dome | ESPN 8:15 pm | 52,689 |\n| 16 | December 15, 1996 | at Kansas City Chiefs | W 24–19 | 9–6 | Arrowhead Stadium | NBC 4:15 pm | 71,136 |\n| 17 | December 22, 1996 | at Cincinnati Bengals | L 24–31 | 9–7 | Cinergy Field | NBC 1:00 pm | 49,389 |\n\n\nQuestion\nwhen was their first loss of the season?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"October 6, 1996"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-6708",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-10337.csv` in your code execution environment.\n\nTable Title: Charleston Southern Buccaneers\nTable Content:\n| Year | FBS Opponent | Result | Opponent's Conference | Opponent's Head Coach | Charleston Southern's Head Coach |\n| --- | --- | --- | --- | --- | --- |\n| 2015 | Alabama Crimson Tide | TBD | SEC | Nick Saban | Jamey Chadwell |\n| 2014 | Georgia Bulldogs | TBD | SEC | Mark Richt | Jamey Chadwell |\n| 2014 | Vanderbilt Commodores | TBD | SEC | Derek Mason | Jamey Chadwell |\n| 2013 | Colorado Buffaloes | L, 43-10 | Pac-12 | Mike MacIntyre | Jamey Chadwell |\n| 2012 | Illinois Fighting Illini | L, 44-0 | Big Ten | Tim Beckman | Jay Mills |\n| 2011 | UCF Knights | L, 62-0 | C-USA | George O'Leary | Jay Mills |\n| 2011 | Florida State Seminoles | L, 62-10 | ACC | Jimbo Fisher | Jay Mills |\n| 2010 | Kentucky Wildcats | L, 49-21 | SEC | Joker Phillips | Jay Mills |\n| 2010 | Hawaii Warriors | L, 66-7 | WAC | Greg McMackin | Jay Mills |\n| 2009 | South Florida Bulls | L, 59-0 | Big East | Jim Leavitt | Jay Mills |\n| 2009 | Florida Gators | L, 62-3 | SEC | Urban Meyer | Jay Mills |\n| 2008 | Miami Redhawks | L, 38-27 | MAC | Don Treadwell | Jay Mills |\n| 2008 | Miami Hurricanes | L, 52-7 | ACC | Randy Shannon | Jay Mills |\n| 2007 | Hawaii Warriors | L, 66-10 | WAC | June Jones | Jay Mills |\n| 2003 | South Florida Bulls | L, 55-7 | Big East | Jim Leavitt | Jay Mills |\n| 2002 | South Florida Bulls | L, 56-6 | Big East | Jim Leavitt | David Dowd |\n\n\nQuestion\nthe fbs opponent faced before the georgia bulldogs?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"Vanderbilt Commodores"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-10337",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-2559.csv` in your code execution environment.\n\nTable Title: 1960 BC Lions season\nTable Content:\n| Week | Date | Opponent | Score | Result | Attendance | Record |\n| --- | --- | --- | --- | --- | --- | --- |\n| 1 | Thursday, Aug 11 | vs. Winnipeg Blue Bombers | 35–21 | Loss | 31,837 | 0–1 |\n| 2 | Monday, Aug 15 | at Edmonton Eskimos | 33–14 | Loss | 17,500 | 0–2 |\n| 3 | Saturday, Aug 20 | vs. Saskatchewan Roughriders | 27–12 | Win | 29,532 | 1–2 |\n| 4 | Monday, Aug 22 | at Calgary Stampeders | 26–19 | Win | n/a | 2–2 |\n| 5 | Monday, Aug 29 | vs. Edmonton Eskimos | 26–0 | Loss | 28,420 | 2–3 |\n| 6 | Thursday, Sept 1 | at Winnipeg Blue Bombers | 19–14 | Loss | 18,297 | 2–4 |\n| 7 | Monday, Sept 5 | at Saskatchewan Roughriders | 31–21 | Win | 14,105 | 3–4 |\n| 8 | Monday, Sept 12 | vs. Calgary Stampeders | 21–21 | Tie | 27,759 | 3–4–1 |\n| 9 | Saturday, Sept 17 | vs. Winnipeg Blue Bombers | 26–14 | Loss | 30,292 | 3–5–1 |\n| 10 | Monday, Sept 19 | at Edmonton Eskimos | 18–10 | Loss | n/a | 3–6–1 |\n| 11 | Saturday, Sept 24 | at Calgary Stampeders | 28–14 | Loss | 13,000 | 3–7–1 |\n| 12 | Thursday, Oct 6 | vs. Edmonton Eskimos | 21–13 | Win | 21,707 | 4–7–1 |\n| 13 | Thursday, Oct 13 | at Winnipeg Blue Bombers | 49–21 | Loss | 16,773 | 4–8–1 |\n| 14 | Saturday, Oct 15 | at Saskatchewan Roughriders | 14–14 | Tie | 7,255 | 4–8–2 |\n| 15 | Saturday, Oct 22 | vs. Calgary Stampeders | 22–10 | Loss | 29,599 | 4–9–2 |\n| 16 | Saturday, Oct 29 | vs. Saskatchewan Roughriders | 38–0 | Win | 21,114 | 5–9–2 |\n\n\nQuestion\nhow many times did a game result in a loss?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"9"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-2559",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-9738.csv` in your code execution environment.\n\nTable Title: Colman Domingo\nTable Content:\n| Genre | Year | Title | Role | Episodes | Notes |\n| --- | --- | --- | --- | --- | --- |\n| Film | 1995 | Timepiece | Khris | | |\n| TV series | 1997 | Nash Bridges | Reggie Harell | \"Dirty Tricks\" | 4 episodes total |\n| Film | 1998 | Around the Fire | Trace | | |\n| Short film | 1999 | King of the Bingo Game | Sonny | | |\n| TV series | 1999 | Nash Bridges | Hassam\\nDesmond Kenner | \"Power Play\"\\n\"Girl Trouble\" | 4 episodes total |\n| Film | 1999 | True Crime | Wally Cartwright | | |\n| TV series | 2000 | Nash Bridges | Trumpet Player | \"Skin Trade\" | 4 episodes total |\n| Film | 2000 | Desi's Looking for a New Girl | Mother | | |\n| Film | 2003 | Kung Phooey! | Roy Lee | | |\n| TV series | 2004 | Law & Order | Ronald Gumer | \"Hands Free\" | 2 episodes total |\n| Film | 2006 | Freedomland | Male Patient | | |\n| TV series | 2006 | Law & Order: Criminal Intent | Sergeant Ev Sides | \"To the Bone\" | 2 episodes total |\n| TV series | 2006 | Law & Order: Trial by Jury | Gus | \"Eros in the Upper Eighties\" | |\n| TV series | 2008 | The Big Gay Sketch Show | | | 8 episodes; 16 episodes total |\n| TV series | 2008 | Law & Order | Donnie | \"Sweetie\" | 2 episodes total |\n| Film | 2008 | Miracle at St. Anna | West Indian Postal Customer | | |\n| TV series | 2009 | Great Performances | Mr. Franklin/Mr. Venus/Joop | \"Passing Strange\" | musical telecast |\n| TV series | 2010 | The Big Gay Sketch Show | | | 8 episodes; 16 episodes total |\n| TV series | 2010 | Law & Order: Criminal Intent | Andre Lanier | \"Love Sick\" | 2 episodes total |\n| Film | 2012 | Lincoln | Private Harold Green | | |\n| Film | 2012 | Red Hook Summer | Blessing Rowe | | |\n| Film | 2013 | 42 | Lawson Bowman | | |\n| Film | 2013 | HairBrained | Finals Moderator | | |\n| Film | 2013 | The Butler | Freddie Fallows | | |\n| Film | 2013 | 400 Boys | Talon | | |\n\n\nQuestion\nwhat was coleman domingo's first tv series?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"Nash Bridges"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-9738",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-12319.csv` in your code execution environment.\n\nTable Title: Patrik Berger\nTable Content:\n| # | Date | Venue | Opponent | Score | Result | Competition |\n| --- | --- | --- | --- | --- | --- | --- |\n| 1. | 6 September 1994 | Bazaly, Ostrava | Malta | 6–1 | 6–1 | Euro 1996 qualifying |\n| 2. | 8 March 1995 | Stadion Za Lužánkami, Brno | Finland | 1–0 | 4–1 | Friendly |\n| 3. | 8 March 1995 | Stadion Za Lužánkami, Brno | Finland | 2–0 | 4–1 | Friendly |\n| 4. | 29 March 1995 | Bazaly, Ostrava | Belarus | 2–0 | 4–2 | Euro 1996 qualifying |\n| 5. | 29 March 1995 | Bazaly, Ostrava | Belarus | 3–1 | 4–2 | Euro 1996 qualifying |\n| 6. | 26 April 1995 | Letná Stadium, Prague | Netherlands | 3–1 | 3–1 | Euro 1996 qualifying |\n| 7. | 7 October 1995 | Dinamo Stadium, Minsk | Belarus | 2–0 | 2–0 | Euro 1996 qualifying |\n| 8. | 15 November 1995 | Letná Stadium, Prague | Luxembourg | 3–0 | 3–0 | Euro 1996 qualifying |\n| 9. | 30 June 1996 | Wembley Stadium, London | Germany | 1–0 | 1–2 | Euro 1996 |\n| 10. | 18 September 1996 | Na Stínadlech, Teplice | Malta | 1–0 | 6–0 | 1998 World Cup qualification |\n| 11. | 18 September 1996 | Na Stínadlech, Teplice | Malta | 3–0 | 6–0 | 1998 World Cup qualification |\n| 12. | 26 February 1997 | Zimní Stadion, Podebrady | Belarus | 4–1 | 4–1 | Friendly |\n| 13. | 14 October 1998 | Na Stínadlech, Teplice | Estonia | 2–0 | 4–1 | Euro 2000 qualifying |\n| 14. | 14 October 1998 | Na Stínadlech, Teplice | Estonia | 3–0 | 4–1 | Euro 2000 qualifying |\n| 15. | 27 March 1999 | Na Stínadlech, Teplice | Lithuania | 2–0 | 2–0 | Euro 2000 qualifying |\n| 16. | 5 June 1999 | Kadrioru Stadium, Tallinn | Estonia | 1–0 | 2–0 | Euro 2000 qualifying |\n| 17. | 8 September 1999 | Na Stínadlech, Teplice | Bosnia and Herzegovina | 2–0 | 3–0 | Euro 2000 qualifying |\n| 18. | 3 June 2000 | Frankenstadion, Nuremberg | Germany | 2–2 | 2–3 | Friendly |\n\n\nQuestion\nwhich team did czech republic score the most goals against?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"Malta"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-12319",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-9887.csv` in your code execution environment.\n\nTable Title: 2008 Clemson Tigers football team\nTable Content:\n| Date | Time | Opponent# | Rank# | Site | TV | Result | Attendance |\n| --- | --- | --- | --- | --- | --- | --- | --- |\n| August 30 | 8:00 PM | vs. Alabama* | #9 | Georgia Dome • Atlanta, GA (Chick-fil-A College Kickoff) | ABC | L 10–34 | 70,097 |\n| September 6 | 3:30 PM | The Citadel* | #22 | Memorial Stadium • Clemson, SC (Military Appreciation Day) | | W 45–17 | 76,794 |\n| September 13 | 12:00 PM | NC State | #23 | Memorial Stadium • Clemson, SC (Textile Bowl/Hall of Fame Day) | Raycom | W 27–9 | 77,071 |\n| September 20 | 1:00 PM | SC State* | #21 | Memorial Stadium • Clemson, SC (Youth Day) | | W 54–0 | 78,607 |\n| September 27 | 12:00 PM | Maryland | #19 | Memorial Stadium • Clemson, SC (Family Weekend) | Raycom | L 17–20 | 81,500 |\n| October 9 | 7:45 PM | at Wake Forest | | BB&T Field • Winston-Salem, NC | ESPN | L 7–12 | 33,988 |\n| October 18 | 12:00 PM | Georgia Tech | | Memorial Stadium • Clemson, SC (IPTAY Day) | ESPN | L 17–21 | 81,500 |\n| November 1 | 3:30 PM | at Boston College | | Alumni Stadium • Chestnut Hill, MA (O'Rourke-McFadden trophy) | ESPNU | W 27–21 | 41,863 |\n| November 8 | 3:30 PM | at #24 Florida State | | Doak Campbell Stadium • Tallahassee, FL | ABC | L 27–41 | 77,013 |\n| November 15 | 12:00 PM | Duke | | Memorial Stadium • Clemson, SC (Homecoming) | Raycom | W 31–7 | 76,217 |\n| November 22 | 12:00 PM | at Virginia | | Scott Stadium • Charlottesville, VA | Raycom | W 13–3 | 51,979 |\n| November 29 | 12:00 PM | South Carolina* | | Memorial Stadium • Clemson, SC (Battle for the Palmetto State/Solid Orange Game) | ESPN2 | W 31–14 | 82,456 |\n| January 1, 2009 | 1:00 PM | vs. Nebraska* | | Jacksonville Municipal Stadium • Jacksonville, FL (Gator Bowl) | CBS | L 21–26 | 67,282 |\n\n\nQuestion\nwhat was the total attendance at the august 30 game against alabama?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"70,097"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-9887",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-5486.csv` in your code execution environment.\n\nTable Title: Patrik Berger\nTable Content:\n| # | Date | Venue | Opponent | Score | Result | Competition |\n| --- | --- | --- | --- | --- | --- | --- |\n| 1. | 6 September 1994 | Bazaly, Ostrava | Malta | 6–1 | 6–1 | Euro 1996 qualifying |\n| 2. | 8 March 1995 | Stadion Za Lužánkami, Brno | Finland | 1–0 | 4–1 | Friendly |\n| 3. | 8 March 1995 | Stadion Za Lužánkami, Brno | Finland | 2–0 | 4–1 | Friendly |\n| 4. | 29 March 1995 | Bazaly, Ostrava | Belarus | 2–0 | 4–2 | Euro 1996 qualifying |\n| 5. | 29 March 1995 | Bazaly, Ostrava | Belarus | 3–1 | 4–2 | Euro 1996 qualifying |\n| 6. | 26 April 1995 | Letná Stadium, Prague | Netherlands | 3–1 | 3–1 | Euro 1996 qualifying |\n| 7. | 7 October 1995 | Dinamo Stadium, Minsk | Belarus | 2–0 | 2–0 | Euro 1996 qualifying |\n| 8. | 15 November 1995 | Letná Stadium, Prague | Luxembourg | 3–0 | 3–0 | Euro 1996 qualifying |\n| 9. | 30 June 1996 | Wembley Stadium, London | Germany | 1–0 | 1–2 | Euro 1996 |\n| 10. | 18 September 1996 | Na Stínadlech, Teplice | Malta | 1–0 | 6–0 | 1998 World Cup qualification |\n| 11. | 18 September 1996 | Na Stínadlech, Teplice | Malta | 3–0 | 6–0 | 1998 World Cup qualification |\n| 12. | 26 February 1997 | Zimní Stadion, Podebrady | Belarus | 4–1 | 4–1 | Friendly |\n| 13. | 14 October 1998 | Na Stínadlech, Teplice | Estonia | 2–0 | 4–1 | Euro 2000 qualifying |\n| 14. | 14 October 1998 | Na Stínadlech, Teplice | Estonia | 3–0 | 4–1 | Euro 2000 qualifying |\n| 15. | 27 March 1999 | Na Stínadlech, Teplice | Lithuania | 2–0 | 2–0 | Euro 2000 qualifying |\n| 16. | 5 June 1999 | Kadrioru Stadium, Tallinn | Estonia | 1–0 | 2–0 | Euro 2000 qualifying |\n| 17. | 8 September 1999 | Na Stínadlech, Teplice | Bosnia and Herzegovina | 2–0 | 3–0 | Euro 2000 qualifying |\n| 18. | 3 June 2000 | Frankenstadion, Nuremberg | Germany | 2–2 | 2–3 | Friendly |\n\n\nQuestion\nwhat opponent is listed last on the table?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"Germany"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-5486",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-9703.csv` in your code execution environment.\n\nTable Title: 2004 African Judo Championships\nTable Content:\n| Rank | Nation | Gold | Silver | Bronze | Total |\n| --- | --- | --- | --- | --- | --- |\n| 1 | Algeria | 8 | 3 | 0 | 11 |\n| 2 | Tunisia | 4 | 3 | 5 | 12 |\n| 3 | Egypt | 3 | 3 | 3 | 9 |\n| 4 | Cameroon | 1 | 1 | 5 | 7 |\n| 5 | Morocco | 0 | 1 | 5 | 5 |\n| 6 | South Africa | 0 | 1 | 1 | 2 |\n| 7 | Angola | 0 | 1 | 0 | 1 |\n| 7 | Burkina Faso | 0 | 1 | 0 | 1 |\n| 7 | Gabon | 0 | 1 | 0 | 1 |\n| 7 | Madagascar | 0 | 1 | 0 | 1 |\n| 11 | Nigeria | 0 | 0 | 3 | 3 |\n| 12 | Senegal | 0 | 0 | 2 | 2 |\n| 13 | Congo Republic | 0 | 0 | 1 | 1 |\n| 13 | Ivory Coast | 0 | 0 | 1 | 1 |\n| 13 | Guinea | 0 | 0 | 1 | 1 |\n| 13 | Niger | 0 | 0 | 1 | 1 |\n\n\nQuestion\nwhat rank did the nation with the most total medals receive?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"2"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-9703",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-9120.csv` in your code execution environment.\n\nTable Title: Atmel AVR\nTable Content:\n| Chip | Flash size | EEPROM | SRAM | Frequency\\n[MHz] | Package |\n| --- | --- | --- | --- | --- | --- |\n| AT90S1200 | 1k | 64 | 0 | 12 | PDIP-20 |\n| AT90S2313 | 2k | 128 | 128 | 10 | PDIP-20 |\n| AT90S/LS2323 | 2k | 128 | 128 | 10 | PDIP-8 |\n| AT90S/LS2343 | 2k | 128 | 128 | 10 | PDIP-8 |\n| AT90S4414 | 4k | 256 | 256 | 8 | PDIP-40 |\n| AT90S/LS4434 | 4k | 256 | 256 | 8 | PDIP-40 |\n| AT90S8515 | 8k | 512 | 512 | 8 | PDIP-40 |\n| AT90S/LS8535 | 8k | 512 | 512 | 8 | PDIP-40 |\n\n\nQuestion\nhow many microcontrollers have 128 eeprom?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"3"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-9120",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-9412.csv` in your code execution environment.\n\nTable Title: Millbrook First Nation\nTable Content:\n| Community | Area | Location | Population | Date established |\n| --- | --- | --- | --- | --- |\n| Beaver Lake 17 | 49.4 hectares (122 acres) | 78.4 km. southeast of Halifax | 23 | March 3, 1867 |\n| Cole Harbour 30 | 18.6 hectares (46 acres) | 9.6 km. east of Halifax | 194 | March 3, 1880 |\n| Millbrook 27 | 302.0 hectares (746 acres) | 8 km. south of Truro | 847 | March 3, 1886 |\n| Sheet Harbour 36 | 32.7 hectares (81 acres) | 91.2 km. northeast of Halifax | 15 | March 3, 1915 |\n| Truro 27A | 16.7 hectares (41 acres) | Joined south of Truro town limit | 0 | March 3, 1904 |\n| Truro 27B | 16.4 hectares (41 acres) | Joined with 27A on south | 0 | March 3, 1907 |\n| Truro 27C | 9.5 hectares (23 acres) | Joined with 27B on south | 0 | March 3, 1909 |\n\n\nQuestion\nwhich community has the least number of acres?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"Truro 27C"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-9412",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-7368.csv` in your code execution environment.\n\nTable Title: Banat\nTable Content:\n| Year | Total | Serbs | Hungarians | Germans | Romanians | Slovaks |\n| --- | --- | --- | --- | --- | --- | --- |\n| 1910 | 566,400 | 229,568 (40.5%) | 108,622 (19.2%) | 125,374 (22.1%) | 73,303 (12.9%) | 16,223 (2,9%) |\n| 1921 | 559,096 | 235,148 (42.1%) | 98,463 (17.6%) | 126,519 (22.6%) | 66,433 (11,9%) | 17,595 (3,2%) |\n| 1931 | 585,579 | 261,123 (44,6%) | 95,867 (16,4%) | 120,541 (20,6%) | 62,365 (10,7%) | 17,900 (2,1%) |\n| 1948 | 601,626 | 358,067 (59,6%) | 110,446 (18,4%) | 17,522 (2,9%) | 55,678 (9,3%) | 20,685 (2,4%) |\n| 1953 | 617,163 | 374,258 (60,6%) | 112,683 (18,4%) | n/a | 55,094 (8,9%) | 21,299 (3,4%) |\n| 1961 | 655,868 | 423,837 (64,6%) | 111,944 (17,1%) | n/a | 54,447 (8,3%) | 22,306 (3,4%) |\n| 1971 | 666,559 | 434,810 (65,2%) | 103,090 (15.5%) | n/a | 49,455 (7,4%) | 22,173 (3,3%) |\n| 1981 | 672,884 | 424,765 (65,7%) | 90,445 (14,0%) | n/a | 43,474 (6,7%) | 21,392 (3,3%) |\n| 1991 | 648,390 | 423,475 (65,1%) | 76,153 (11.7%) | n/a | 35,935 (5,5%) | 19,903 (3.1%) |\n| 2002 | 665,397 | 477,890 (71.8%) | 63,047 (9.5%) | 908 (0,1%) | 27,661 (4,1%) | 17,994 (2,7%) |\n\n\nQuestion\nwhat was the last year that romanians were 5,5%?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"1991"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-7368",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-6739.csv` in your code execution environment.\n\nTable Title: Kazakhstan\nTable Content:\n| Ethnicity | Islam | Christianity | Judaism | Buddhism | Other | None / Atheism | n/a |\n| --- | --- | --- | --- | --- | --- | --- | --- |\n| TOTALS | 70.20% | 26.32% | 0.03% | 0.09% | 0.02% | 2.82% | 0.51% |\n| Kazakh | 98.34% | 0.39% | 0.02% | 0.01% | 0.02% | 0.98% | 0.26% |\n| Russian | 1.43% | 91.64% | 0.04% | 0.02% | 0.03% | 6.09% | 0.75% |\n| Uzbek | 99.05% | 0.39% | 0.01% | 0.01% | 0.02% | 0.37% | 0.16% |\n| Ukrainian | 0.94% | 90.74% | 0.03% | 0.01% | 0.02% | 7.31% | 0.94% |\n| Uyghur | 98.35% | 0.51% | 0.02% | 0.01% | 0.03% | 0.61% | 0.47% |\n| Tatar | 79.57% | 10.24% | 0.02% | 0.03% | 0.06% | 8.11% | 1.97% |\n| German | 1.58% | 81.59% | 0.05% | 0.04% | 0.11% | 13.96% | 2.68% |\n| Korean | 5.24% | 49.35% | 0.21% | 11.40% | 0.14% | 28.51% | 5.16% |\n| Turkish | 99.13% | 0.30% | 0.01% | 0.01% | 0.02% | 0.33% | 0.21% |\n| Azeri | 94.81% | 2.51% | 0.02% | 0.02% | 0.03% | 1.86% | 0.76% |\n| Belorussian | 0.79% | 90.16% | 0.04% | 0.01% | 0.03% | 7.82% | 1.15% |\n| Dungan | 98.93% | 0.37% | 0.01% | 0.03% | 0.04% | 0.34% | 0.28% |\n| Kurdish | 98.28% | 0.53% | 0.03% | 0.02% | 0.02% | 0.74% | 0.38% |\n| Tajik | 97.78% | 0.91% | 0.01% | 0.02% | 0.08% | 0.85% | 0.35% |\n| Polish | 0.69% | 90.07% | 0.04% | 0.01% | 0.13% | 7.30% | 1.76% |\n| Chechen | 93.69% | 2.99% | 0.02% | 0.01% | 0.05% | 2.08% | 1.16% |\n| Kyrgyz | 96.67% | 0.89% | 0.03% | 0.03% | 0.02% | 1.51% | 0.86% |\n| Others | 34.69% | 52.32% | 0.82% | 0.91% | 0.13% | 8.44% | 2.69% |\n\n\nQuestion\nwhich ethnicity is first on the chart\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"Kazakh"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-6739",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-6900.csv` in your code execution environment.\n\nTable Title: Senior Advisor to the President of the United States\nTable Content:\n| Name | State of residence | Took office | Left office | President served under |\n| --- | --- | --- | --- | --- |\n| Karl Rove | Texas | January 21, 2001 | August 31, 2007 | George W. Bush |\n| Barry Steven Jackson | Ohio | September 1, 2007 | January 19, 2009 | George W. Bush |\n| Pete Rouse | South Dakota | January 20, 2009 | October 1, 2010 | Barack Obama |\n| David Axelrod | Illinois | January 20, 2009 | January 10, 2011 | Barack Obama |\n| David Plouffe | Delaware | January 10, 2011 | January 25, 2013 | Barack Obama |\n| Valerie Jarrett | Illinois | January 20, 2009 | present | Barack Obama |\n| Daniel Pfeiffer | Delaware | January 25, 2013 | present | Barack Obama |\n\n\nQuestion\nwho was the last senior advisor to be named in 2013?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"Daniel Pfeiffer"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-6900",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-8232.csv` in your code execution environment.\n\nTable Title: Bears–Lions rivalry\nTable Content:\n| Year | Date | Winner | Result | Loser | Attendance | Location |\n| --- | --- | --- | --- | --- | --- | --- |\n| 1980 | Oct 19 | Chicago Bears | 24-7 | Detroit Lions | | Chicago, Illinois |\n| 1980 | Nov 27 | Chicago Bears | 23-17 (OT) | Detroit Lions | 58,508 | Pontiac, Michigan |\n| 1981 | Oct 19 | Detroit Lions | 48-17 | Chicago Bears | 75,397 | Pontiac, Michigan |\n| 1981 | Nov 22 | Detroit Lions | 23-7 | Chicago Bears | 71,273 | Chicago, Illinois |\n| 1982 | Sept 12 | Detroit Lions | 17-10 | Chicago Bears | 50,082 | Pontiac, Michigan |\n| 1982 | Nov 21 | Chicago Bears | 20-17 | Detroit Lions | 46,783 | Chicago, Illinois |\n| 1983 | Oct 16 | Detroit Lions | 31–17 | Chicago Bears | 66,709 | Pontiac, Michigan |\n| 1983 | Oct 30 | Detroit Lions | 38-17 | Chicago Bears | 58,764 | Chicago, Illinois |\n| 1984 | Nov 18 | Chicago Bears | 16-14 | Detroit Lions | 54,911 | Chicago, Illinois |\n| 1984 | Dec 16 | Chicago Bears | 30-13 | Detroit Lions | 53,252 | Pontiac, Michigan |\n| 1985 | Nov 10 | Chicago Bears | 24-3 | Detroit Lions | 53,467 | Chicago, Illinois |\n| 1985 | Dec 22 | Chicago Bears | 37-17 | Detroit Lions | 74,042 | Pontiac, Michigan |\n| 1986 | Oct 26 | Chicago Bears | 13-7 | Detroit Lions | 62,064 | Chicago, Illinois |\n| 1986 | Dec 15 | Chicago Bears | 16-13 | Detroit Lions | 75,602 | Pontiac, Michigan |\n| 1987 | Nov 22 | Chicago Bears | 30-10 | Detroit Lions | 63,357 | Chicago, Illinois |\n| 1988 | Oct 9 | Chicago Bears | 24–7 | Detroit Lions | 64,526 | Pontiac, Michigan |\n| 1988 | Dec 11 | Chicago Bears | 13-12 | Detroit Lions | 55,010 | Chicago, Illinois |\n| 1989 | Sept 24 | Chicago Bears | 47-27 | Detroit Lions | 71,418 | Pontiac, Michigan |\n| 1989 | Dec 10 | Detroit Lions | 27-17 | Chicago Bears | 52,650 | Chicago, Illinois |\n\n\nQuestion\nwhich date was attended by the most people?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"December 15, 1986"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-8232",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-7988.csv` in your code execution environment.\n\nTable Title: Laredo Heat\nTable Content:\n| Year | Division | League | Regular Season | Playoffs | Open Cup |\n| --- | --- | --- | --- | --- | --- |\n| 2004 | 4 | USL PDL | 8th, Mid South | Did not qualify | Did not qualify |\n| 2005 | 4 | USL PDL | 2nd, Mid South | Conference Finals | Did not qualify |\n| 2006 | 4 | USL PDL | 1st, Mid South | National Final | 1st Round |\n| 2007 | 4 | USL PDL | 1st, Mid South | Champions | Did not qualify |\n| 2008 | 4 | USL PDL | 2nd, Mid South | National Final | Did not qualify |\n| 2009 | 4 | USL PDL | 1st, Mid South | Conference Finals | Did not qualify |\n| 2010 | 4 | USL PDL | 1st, Mid South | Conference Semi-Final | Did not qualify |\n| 2011 | 4 | USL PDL | 1st, Mid South | National Final | Did not qualify |\n| 2012 | 4 | USL PDL | 1st, Mid South | Conference Semi-Final | 2nd Round |\n| 2013 | 4 | USL PDL | 2nd, Mid South | Conference Finals | 2nd Round |\n\n\nQuestion\nlast year not to make the open cup\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"2011"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-7988",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-9772.csv` in your code execution environment.\n\nTable Title: Noosa Festival of Surfing\nTable Content:\n| Event | Winner | Second | Third | Fourth |\n| --- | --- | --- | --- | --- |\n| 2008 Telstra Men's Pro | Josh Constable | Seb Wilson | Grant Thomas | Jackson Close |\n| 2008 Banana Boat Women's Pro | Jennifer Smith | Janna Irons | Justin Dupont | Selby Riddle |\n| 2007 Odyssey 20/20 Men's Pro | Josh Constable | Harrison Roach | Jackson Close | Harley Ingleby |\n| 2007 Banana Boat Women's Pro | Chelsea Wlliams | Lisa Nicholson | Isabelle Braley | Crystal Dzigas |\n| 2009 Golden Breed Mens Pro LQS | Taylor Jensen | Josh Constable | Christian Wach | Harrison Roach |\n| 2009 Banana Boat Women's Pro | Chelsea Wlliams | Melissa Combo | IsaSelby Riddle | Isabelle Brayley |\n| 2010 Golden Breed Mens Pro LQS | Taylor Jensen | Josh Constable | Christian Wach | Jackson Close |\n| 2010 The Back Beach Women’s 1 Star Pro | Crystal Dzigas | Isabele Braly | Leah Dawson | Rosie Locke |\n\n\nQuestion\nname each of the years that taylor jensen was winner.\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"2009",
"2010"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-9772",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-6501.csv` in your code execution environment.\n\nTable Title: 1984 Detroit Tigers season\nTable Content:\n| # | Date | Opponent | Score | Win | Loss | Save | Crowd | Record |\n| --- | --- | --- | --- | --- | --- | --- | --- | --- |\n| 1 | Apr 3 | @MIN | W 8-1 | Morris | Williams | | 34,381 | 1-0 |\n| 2 | Apr 5 | @MIN | W 7-3 | Petry | Viola | | 8,373 | 2-0 |\n| 3 | Apr 6 | @CHW | W 3-2 | Wilcox | Dotson | Hernandez | 42,692 | 3-0 |\n| 4 | Apr 7 | @CHW | W 4-0 | Morris | Bannister | | 24,616 | 4-0 |\n| 5 | Apr 8 | @CHW | W 7-3 | Lopez | Seaver | | 20,478 | 5-0 |\n| 6 | Apr 10 | TEX | W 5-1 | Petry | Stewart | | 51,238 | 6-0 |\n| 7 | Apr 12 | TEX | W 9-4 | Morris | Tanana | | 19,154 | 7-0 |\n| 8 | Apr 13 | @BOS | W 13-9 | Bair | Hurst | | 35,179 | 8-0 |\n| 9 | Apr 18 | KCR | W 4-3 (10) | Hernandez | Beckwith | | 12,310 | 9-0 |\n| 10 | Apr 19 | KCR | L 2-5 | Saberhagen | Petry | Quisenberry | 12,100 | 9-1 |\n| 11 | Apr 20 | CHW | W 3-2 | Lopez | Reed | | 33,554 | 10-1 |\n| 12 | Apr 21 | CHW | W 4-1 | Rozema | Hoyt | Bair | 34,395 | 11-1 |\n| 13 | Apr 22 | CHW | W 9-1 | Berenguer | Brennan | | 10,603 | 12-1 |\n| 14 | Apr 24 | MIN | W 6-5 | Morris | Davis | | N/A | 13-1 |\n| 15 | Apr 24 | MIN | W 4-3 | Abbott | Viola | Lopez | 20,315 | 14-1 |\n| 16 | Apr 25 | @TEX | W 9-4 | Wilcox | Stewart | Hernandez | 25,883 | 15-1 |\n| 17 | Apr 26 | @TEX | W 7-5 | Bair | Tanana | Lopez | 13,559 | 16-1 |\n| 18 | Apr 27 | CLE | L 4-8 (19) | Aponte | Abbott | | 34,112 | 16-2 |\n| 19 | Apr 28 | CLE | W 6-2 | Morris | Behenna | | 28,253 | 17-2 |\n| 20 | Apr 29 | CLE | W 6-1 | Petry | Spillner | | 24,853 | 18-2 |\n\n\nQuestion\nbefore april 7, how many times had morris won?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"1"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-6501",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-8905.csv` in your code execution environment.\n\nTable Title: Sri Lankan IDP numbers, February and March 2009\nTable Content:\n| Camp | Div Sec\\nDivision | Admin\\nDistrict | 9 Feb | 11 Feb | 16 Feb | 25 Feb | 27 Feb | 2 Mar | 4 Mar | 6 Mar | 11 Mar | 13 Mar | 24 Mar | 30 Mar | 31 Mar |\n| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |\n| Menik Farm Zone 0\\n(Kathirkamar Village) | Vengalachedikulam | VAV | | | 1,353 | 2,768 | 2,791 | 2,791 | 3,344 | 4,212 | 4,679 | 4,679 | 6,459 | 6,459 | 6,459 |\n| Menik Farm Zone 3\\n(Arunachchalam Village) | Vengalachedikulam | VAV | | | | | | | | | | | 6,729 | 11,881 | 11,881 |\n| Ariviththodam Sivanantha Vidyalayam, Menik Farm | Vengalachedikulam | VAV | 581 | 581 | 581 | 581 | 584 | 584 | 585 | 585 | 583 | 583 | 601 | 601 | 601 |\n| Cheddikulam Maha Vidyalayam | Vengalachedikulam | VAV | 1,624 | 1,624 | 1,624 | 1,965 | 1,965 | 1,965 | 1,996 | 1,996 | 2,001 | 2,001 | 1,995 | 1,995 | 1,995 |\n| Gamini Maha Vidyalayam | Vavuniya | VAV | 1,306 | 1,306 | 2,017 | 1,618 | 1,618 | 1,618 | 1,263 | 1,307 | 1,382 | 1,382 | 1,471 | 1,487 | 1,487 |\n| Kovilkulam Hindu College | Vavuniya | VAV | | 1,066 | 1,056 | 1,081 | 1,143 | 1,143 | 1,130 | 1,130 | 1,139 | 1,139 | 1,167 | 1,166 | 1,166 |\n| Nelukkulam Kalaimahal Maha Vidyalayam | Vavuniya | VAV | 2,829 | 2,829 | 2,829 | 3,071 | 3,084 | 3,084 | 3,084 | 3,068 | 3,068 | 3,082 | 3,101 | 3,070 | 3,070 |\n| Nelukkulam Technical College | Vavuniya | VAV | | 1,527 | 2,374 | 2,913 | 2,913 | 2,913 | 2,730 | 2,801 | 2,801 | 2,968 | 3,020 | 3,032 | 3,032 |\n| Pampamadhu Hostel School | Vavuniya | VAV | 2,779 | 2,779 | 4,181 | 4,131 | 4,131 | 4,131 | 4,131 | 4,108 | 4,108 | 4,108 | 4,178 | 4,178 | 4,178 |\n| Ponthoodam Government Tamil Mixed School | Vavuniya | VAV | 1,351 | 1,351 | 1,351 | 1,365 | 1,310 | 1,310 | 1,310 | 1,308 | 1,308 | 1,320 | 1,327 | 1,319 | 1,319 |\n| Poonathoddam College of Education | Vavuniya | VAV | 4,406 | 4,406 | 5,302 | 5,302 | 5,009 | 5,009 | 5,009 | 5,036 | 5,036 | 5,119 | 5,236 | 5,243 | 5,243 |\n| Sivapirakasa Ladies College, Vavuniya | Vavuniya | VAV | | | | | 2,257 | 2,257 | 2,346 | 2,390 | 2,612 | 2,612 | 3,020 | 2,973 | 2,973 |\n| Vavuniya Muslim Maha Vidyalayam | Vavuniya | VAV | | 1,333 | 1,767 | 1,767 | | | | | | | 1,360 | 1,394 | 1,394 |\n| Vavuniya Tamil Maha Vidyalayam (Primary) | Vavuniya | VAV | | | 1,916 | 1,929 | 1,390 | 1,390 | | 1,153 | 1,153 | 1,153 | 1,196 | 1,168 | 1,168 |\n| Vavuniya Tamil Maha Vidyalayam (Senior) | Vavuniya | VAV | | 1,200 | 4,390 | 3,838 | 4,085 | 4,085 | 5,475 | 3,540 | 3,390 | 3,390 | 3,963 | 3,969 | 3,969 |\n| Kaithady Ayurvedic University Hostel | Thenmarachchi | JAF | | | | | | | | | 647 | 647 | 1,073 | 1,073 | 1,083 |\n| Kaithady Hindu Children Home | Thenmarachchi | JAF | | | | | | | | 184 | 184 | 184 | 667 | 667 | 668 |\n| Kaithady Palmyra Research Institute 1 | Thenmarachchi | JAF | | | | | | | | | | | | | 357 |\n| Kodikamam Government Tamil Mixed School | Thenmarachchi | JAF | | | 411 | 424 | 503 | 503 | 655 | 711 | 706 | 706 | 1,006 | 1,006 | 1,006 |\n| Kopay Teacher Training College | Valikamam East | JAF | | | 444 | 438 | 464 | 464 | 463 | 463 | 463 | 463 | 464 | 464 | 464 |\n| Murusivil Roman Catholic Tamil Mixed School | Thenmarachchi | JAF | | | 1,139 | 1,140 | 1,139 | 1,139 | 1,140 | 1,140 | 1,313 | 1,313 | 1,691 | 1,691 | 1,786 |\n| Thirunagar Old Court House | Jaffna | JAF | | | 73 | 74 | 49 | 49 | 108 | 108 | 113 | 113 | 114 | 114 | 140 |\n| Illupaikkulam | Mannar Town | MAN | | | 52 | 52 | 52 | 52 | 52 | 52 | 52 | 52 | 52 | 52 | 52 |\n| Kalimoddai | Nanaddan | MAN | | | 445 | 445 | 445 | 445 | 445 | 454 | 454 | 454 | 454 | 454 | 454 |\n| Mannar District General Hospital | Mannar Town | MAN | | | | | 279 | 279 | 279 | 279 | 279 | 279 | 279 | 279 | 279 |\n| Sirukandal | Nanaddan | MAN | | | 357 | 357 | 357 | 357 | 357 | 350 | 350 | 350 | 350 | 350 | 350 |\n| Kantale Base Hospital | Kantale | TRI | | | | 84 | 160 | 127 | 127 | 127 | 80 | 80 | 57 | 57 | 57 |\n| Methodist School | | TRI | | | | 83 | 52 | 26 | 32 | 126 | 40 | 40 | | | |\n| Pulmoddai Field Hospital | Kuchaveli | TRI | | | | | | | | | | | 79 | 79 | 129 |\n| Thampalakamam Peripheral Unit | Thampalakamam | TRI | | | | 33 | 76 | 79 | 77 | 77 | 75 | 75 | 35 | 35 | 35 |\n| Trincomalee General Hospital | Town & Gravets | TRI | | | | 678 | 544 | 512 | 417 | 492 | 426 | 426 | 374 | 374 | 296 |\n| Colombo National Hospital | | COL | | | | 28 | 75 | 106 | 106 | 106 | 97 | 97 | 186 | 186 | 118 |\n| Padaviya Base Hospital | | ANU | | | | | | | | | | | 145 | 145 | 223 |\n| Polonnaruwa General Hospital | | POL | | | | 86 | 171 | 171 | 171 | 171 | 171 | 171 | 100 | 100 | 92 |\n| Total | Total | Total | 14,876 | 20,002 | 33,662 | 36,251 | 36,646 | 36,589 | 36,832 | 37,474 | 38,710 | 38,986 | 51,949 | 57,061 | 57,524 |\n\n\nQuestion\nwhat division comes after jaffna\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"Mannar Town"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-8905",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-12293.csv` in your code execution environment.\n\nTable Title: Supreme Court of Puerto Rico\nTable Content:\n| Name | Rank | Age | Years until mandatory retirement | Appointed by | Year appointed |\n| --- | --- | --- | --- | --- | --- |\n| Anabelle Rodríguez | Associate Justice | 64 | 6 years | Sila María Calderón | 2004 |\n| Edgardo Rivera Garcia | Associate Justice | 59 | 11 years | Luis Fortuño | 2010 |\n| Erick Kolthoff Caraballo | Associate Justice | 53 | 17 years | Luis Fortuño | 2009 |\n| Federico Hernández Denton | Chief Justice | 70 | 0 year | Sila María Calderón | 2004 |\n| Liana Fiol Matta | Associate Justice | 68 | 2 years | Sila María Calderón | 2004 |\n| Luis Estrella Martínez | Associate Justice | 43 | 27 years | Luis Fortuño | 2011 |\n| Mildred Pabón Charneco | Associate Justice | 57 | 13 years | Luis Fortuño | 2009 |\n| Rafael Martínez Torres | Associate Justice | 55 | 15 years | Luis Fortuño | 2009 |\n| Roberto Feliberti Cintrón | Associate Justice | 51 | 19 years | Luis Fortuño | 2011 |\n\n\nQuestion\nwho was the only associate justice appointed in 2010?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"Edgardo Rivera Garcia"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-12293",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-4578.csv` in your code execution environment.\n\nTable Title: 2013 BWF Super Series\nTable Content:\n| Tour | Official title | Venue | City | Date\\nStart | Date\\nFinish | Prize money\\nUSD | Report |\n| --- | --- | --- | --- | --- | --- | --- | --- |\n| 1 | Korea Open Super Series Premier | Seoul National University Gymnasium | Seoul | January 8 | January 13 | 1,000,000 | Report |\n| 2 | Malaysia Open Super Series | Bukit Kiara Sports Complex | Kuala Lumpur | January 15 | January 20 | 400,000 | Report |\n| 3 | All England Super Series Premier | National Indoor Arena | Birmingham | March 5 | March 10 | 400,000 | Report |\n| 4 | India Super Series | Siri Fort Indoor Stadium | New Delhi | April 23 | April 28 | 200,000 | Report |\n| 5 | Indonesia Super Series Premier | Istora Senayan | Jakarta | June 11 | June 16 | 700,000 | Report |\n| 6 | Singapore Super Series | Singapore Indoor Stadium | Singapore | June 18 | June 23 | 200,000 | Report |\n| 7 | China Masters Super Series | Olympic Sports Centre Xincheng Gymnasium | Changzhou | September 10 | September 15 | 250,000 | Report |\n| 8 | Japan Super Series | Yoyogi National Gymnasium | Tokyo | September 17 | September 22 | 200,000 | Report |\n| 9 | Denmark Super Series Premier | Odense Idrætspark | Odense | October 15 | October 20 | 400,000 | Report |\n| 10 | French Super Series | Stade Pierre de Coubertin | Paris | October 22 | October 27 | 200,000 | Report |\n| 11 | China Open Super Series Premier | Yuanshen Sports Centre Stadium | Shanghai | November 12 | November 17 | 350,000 | Report |\n| 12 | Hong Kong Super Series | Hong Kong Coliseum | Kowloon | November 19 | November 24 | 250,000 | Report |\n| 13 | Super Series Masters Finals | Kuala Lumpur Badminton Stadium | Kuala Lumpur | December 11 | December 15 | 500,000 | Report |\n\n\nQuestion\nwhat is the total prize payout for all 13 series?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"5050000"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-4578",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-11297.csv` in your code execution environment.\n\nTable Title: Pope County, Arkansas\nTable Content:\n| Township | FIPS | Population\\ncenter | Population | Population\\ndensity\\n(/mi²) | Population\\ndensity\\n(/km²) | Land area\\n(mi²) | Land area\\n(km²) | Water area\\n(mi²) | Water area\\n(km²) | Geographic coordinates |\n| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |\n| Bayliss | 90159 | | 708 | 24.6 | 9.5 | 28.81 | 74.62 | 0.0979 | 0.2536 | 35°24′10″N 93°14′06″W / 35.40278°N 93.23500°W |\n| Burnett | 90558 | | 452 | 20.9 | 8.1 | 21.65 | 56.07 | 0.1051 | 0.2722 | 35°19′10″N 92°52′33″W / 35.31944°N 92.87583°W |\n| Center | 90735 | | 515 | 36.8 | 14.2 | 13.99 | 36.23 | 0.0339 | 0.0878 | 35°24′20″N 92°57′16″W / 35.40556°N 92.95444°W |\n| Clark | 90813 | London | 2969 | 115.3 | 44.6 | 25.73 | 66.64 | 6.0444 | 15.6549 | 35°19′45″N 93°14′46″W / 35.32917°N 93.24611°W |\n| Convenience | 90921 | | 933 | 50.4 | 19.4 | 18.53 | 47.99 | 0.0942 | 0.2440 | 35°20′00″N 92°56′41″W / 35.33333°N 92.94472°W |\n| Dover | 91134 | Dover | 5277 | 119.1 | 46.0 | 44.29 | 114.7 | 0.3637 | 0.9420 | 35°23′30″N 93°07′01″W / 35.39167°N 93.11694°W |\n| Freeman | 91377 | | 98 | 0.8 | 0.3 | 119.78 | 310.2 | 0.0000 | 0.0000 | 35°39′10″N 93°04′06″W / 35.65278°N 93.06833°W |\n| Galla | 91407 | Pottsville | 3523 | 88.7 | 34.3 | 39.71 | 102.8 | 1.8410 | 4.7682 | 35°13′15″N 93°02′46″W / 35.22083°N 93.04611°W |\n| Griffin | 91536 | | 901 | 26.5 | 10.2 | 33.96 | 87.96 | 0.1106 | 0.2865 | 35°25′30″N 92°52′36″W / 35.42500°N 92.87667°W |\n| Gum Log | 91560 | | 1420 | 71.6 | 27.6 | 19.84 | 51.39 | 0.0142 | 0.0368 | 35°16′30″N 92°59′51″W / 35.27500°N 92.99750°W |\n| Illinois | 91812 | Russellville | 25841 | 540.9 | 208.9 | 47.77 | 123.7 | 6.6022 | 17.0996 | 35°17′00″N 93°07′46″W / 35.28333°N 93.12944°W |\n| Jackson | 91875 | Hector | 1191 | 11.5 | 4.4 | 103.72 | 268.6 | 0.0505 | 0.1308 | 35°29′20″N 92°57′01″W / 35.48889°N 92.95028°W |\n| Liberty | 92181 | | 805 | 14.2 | 5.5 | 56.64 | 146.7 | 0.0028 | 0.0073 | 35°29′40″N 93°03′16″W / 35.49444°N 93.05444°W |\n| Martin | 92415 | | 1482 | 23.7 | 9.2 | 62.46 | 161.8 | 0.3931 | 1.0181 | 35°28′25″N 93°10′06″W / 35.47361°N 93.16833°W |\n| Moreland | 92553 | | 700 | 52.2 | 20.2 | 13.40 | 34.71 | 0.0683 | 0.1769 | 35°21′30″N 92°59′46″W / 35.35833°N 92.99611°W |\n| Phoenix | 92871 | | 334 | 26.7 | 10.3 | 12.51 | 32.40 | 0.0000 | 0.0000 | 35°24′30″N 93°00′31″W / 35.40833°N 93.00861°W |\n| Smyrna | 93420 | | 173 | 2.4 | 0.9 | 70.69 | 183.1 | 0.0218 | 0.0565 | 35°38′10″N 92°53′46″W / 35.63611°N 92.89611°W |\n| Valley | 93765 | | 2776 | 125.7 | 48.5 | 22.09 | 57.21 | 0.0144 | 0.0373 | 35°20′05″N 93°02′46″W / 35.33472°N 93.04611°W |\n| Wilson | 94089 | Atkins | 4371 | 77.6 | 30.0 | 56.32 | 145.9 | 3.0305 | 7.8490 | 35°13′30″N 92°55′01″W / 35.22500°N 92.91694°W |\n\n\nQuestion\nwhich township in pope county, arkansas has the least amount of water area?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"Freeman",
"Phoenix"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-11297",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-8816.csv` in your code execution environment.\n\nTable Title: Athletics at the 1932 Summer Olympics – Men's 50 kilometres walk\nTable Content:\n| Rank | Name | Nationality | Time (hand) | Notes |\n| --- | --- | --- | --- | --- |\n| | Tommy Green | Great Britain | 4:50:10 | OR |\n| | Jānis Daliņš | Latvia | 4:57:20 | |\n| | Ugo Frigerio | Italy | 4:59:06 | |\n| 4 | Karl Hähnel | Germany | 5:06:06 | |\n| 5 | Ettore Rivolta | Italy | 5:07:39 | |\n| 6 | Paul Sievert | Germany | 5:16:41 | |\n| 7 | Henri Quintric | France | 5:27:25 | |\n| 8 | Ernie Crosbie | United States | 5:28:02 | |\n| 9 | Bill Chisholm | United States | 5:51:00 | |\n| 10 | Alfred Maasik | Estonia | 6:19:00 | |\n| | Henry Cieman | Canada | | DNF |\n| | John Moralis | Greece | | DNF |\n| | Francesco Pretti | Italy | | DNF |\n| | Arthur Tell Schwab | Switzerland | | DNF |\n| | Harry Hinkel | United States | | DNF |\n\n\nQuestion\nhow many times did the united states rank?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"2"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-8816",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-9355.csv` in your code execution environment.\n\nTable Title: 1983 NHL Entry Draft\nTable Content:\n| Pick # | Player | Position | Nationality | NHL team | College/junior/club team |\n| --- | --- | --- | --- | --- | --- |\n| 123 | Paul Ames | Defence | United States | Pittsburgh Penguins | Billerica High School (USHS-MA) |\n| 124 | Joe Reekie | Defence | Canada | Hartford Whalers | North Bay Centennials (OHL) |\n| 125 | Greg Evtushevski | Right Wing | Canada | New Jersey Devils | Kamloops Blazers (WHL) |\n| 126 | Bob Pierson | Left Wing | Canada | Detroit Red Wings | London Knights (OHL) |\n| 127 | Tim Burgess | Defence | Canada | Los Angeles Kings | Oshawa Generals (OHL) |\n| 128 | Cam Plante | Defence | Canada | Toronto Maple Leafs | Brandon Wheat Kings (WHL) |\n| 129 | Iain Duncan | Left Wing | Canada | Winnipeg Jets | North York Rangers (OJHL) |\n| 130 | Terry Maki | Left Wing | Canada | Vancouver Canucks | Brantford Alexanders (OHL) |\n| 131 | Jeff Hogg | Goaltender | Canada | Calgary Flames | Oshawa Generals (OHL) |\n| 132 | Craig Mack | Defence | United States | Quebec Nordiques | East Grand Forks High School (USHS-MN) |\n| 133 | Steve Orth | Centre | United States | New York Rangers | St. Cloud State University (WCHA) |\n| 134 | Christian Ruuttu | Centre | Finland | Buffalo Sabres | Pori (Finland) |\n| 135 | Dwaine Hutton | Centre | Canada | Washington Capitals | Kelowna Wings (WHL) |\n| 136 | Sean Toomey | Left Wing | United States | Minnesota North Stars | Cretin-Derham Hall High School (USHS-MN) |\n| 137 | Jim Sprenger | Defence | United States | New York Islanders | Cloquet High School (USHS-MN) |\n| 138 | Vladislav Tretiak | Goaltender | Soviet Union | Montreal Canadiens | Moscow CSKA (USSR) |\n| 139 | Scott Birnie | Rignt wing | Canada | Chicago Black Hawks | Cornwall Royals (OHL) |\n| 140 | Dale Derkatch | Centre | Canada | Edmonton Oilers | Regina Pats (WHL) |\n| 141 | Bob Mormina | Right Wing | Canada | Philadelphia Flyers | Longueuil Chevaliers (QMJHL) |\n| 142 | Ian Armstrong | Defence | Canada | Boston Bruins | Peterborough Petes (OHL) |\n\n\nQuestion\nwas pick #137's nationality united states, soviet union, or canada?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"United States"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-9355",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-3401.csv` in your code execution environment.\n\nTable Title: 1997 Brisbane Broncos season\nTable Content:\n| Round | Opponent | Result | Bro. | Opp. | Date | Venue | Crowd | Position |\n| --- | --- | --- | --- | --- | --- | --- | --- | --- |\n| 1 | Auckland Warriors | Win | 14 | 2 | 1 Mar | ANZ Stadium | 42,361 | 2/10 |\n| 2 | Adelaide Rams | Win | 28 | 12 | 9 Mar | ANZ Stadium | 16,279 | 1/10 |\n| 3 | Perth Reds | Win | 26 | 16 | 16 Mar | WACA Ground | 11,109 | 2/10 |\n| 4 | Canberra Raiders | Win | 24 | 8 | 24 Mar | ANZ Stadium | 26,103 | 2/10 |\n| 5 | Cronulla Sharks | Win | 14 | 12 | 31 Mar | ANZ Stadium | 17,294 | 1/10 |\n| 6 | North Queensland Cowboys | Win | 42 | 16 | 5 Apr | Stockland Stadium | 30,122 | 1/10 |\n| 7 | Adelaide Rams | Win | 20 | 10 | 13 Apr | Adelaide Oval | 17,633 | 1/10 |\n| 8 | Penrith Panthers | Loss | 26 | 27 | 21 Apr | Penrith Football Stadium | 12,648 | 1/10 |\n| 9 | Canterbury Bulldogs | Win | 34 | 16 | 28 Apr | ANZ Stadium | 17,921 | 1/10 |\n| 10 | Auckland Warriors | Win | 34 | 18 | 4 May | Mt Smart Stadium | 16,471 | 1/10 |\n| 11 | North Queensland Cowboys | Draw | 20 | 20 | 25 May | ANZ Stadium | 14,167 | 1/10 |\n| 12 | Hunter Mariners | Loss | 6 | 24 | 30 May | Topper Stadium | 7,124 | 1/10 |\n| 13 | Cronulla Sharks | Loss | 4 | 32 | 30 Jun | Endeavour Field | 12,240 | 1/10 |\n| 14 | Perth Reds | Win | 50 | 14 | 6 Jul | ANZ Stadium | 11,806 | 1/10 |\n| 15 | Canberra Raiders | Win | 19 | 4 | 14 Jul | Bruce Stadium | 13,067 | 1/10 |\n| 16 | Penrith Panthers | Win | 54 | 12 | 10 Aug | ANZ Stadium | 13,921 | 1/10 |\n| 17 | Hunter Mariners | Win | 34 | 16 | 18 Aug | ANZ Stadium | 13,830 | 1/10 |\n| 18 | Canterbury Bulldogs | Win | 32 | 24 | 25 Aug | Belmore Oval | 9,781 | 1/10 |\n| Semi Final | Cronulla Sharks | Win | 34 | 2 | 6 Sep | Stockland Stadium | 26,256 | |\n| GRAND FINAL | Cronulla Sharks | Win | 26 | 8 | 20 Sep | ANZ Stadium | 58,912 | |\n\n\nQuestion\nin how many games was the crowd above 20,000 people?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"5 games"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-3401",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-1303.csv` in your code execution environment.\n\nTable Title: Cello (web browser)\nTable Content:\n| Version | Date | Development cycle | Size (in kb) | Download | Notes |\n| --- | --- | --- | --- | --- | --- |\n| 0.1 | 9 June 1993 | Beta | ? | evolt | |\n| 0.2 | 14 June 1993 | Beta | ? | ? | Changelog |\n| 0.3 | 16 June 1993 | Beta | ? | ? | Changelog |\n| 0.4 | 18 June 1993 | Beta | ? | ? | Changelog |\n| 0.5 | 24 June 1993 | Beta | ? | ? | Changelog |\n| 0.6 | 30 June 1993 | Beta | ? | ? | changelog |\n| 0.8 | 5 November 1993 | Beta | N/A | N/A | Changelog (Distinct version discontinued) |\n| 0.9 | 12 November 1993 | Beta-pre | ? | [2] | |\n| 0.9 | 16 November 1993 | Beta | ? | [3] | Changelog |\n| 0.9 | 22 November 1993 | WINSOCK alpha r9.2 | ? | [4] | |\n| 1.0 | 17 February 1994 | Release | ? | evolt | |\n| 1.01 | ? | Release | ? | ? | |\n| 1.01a | 17 March 1994 | release | 521 | [5], evolt | Changelog |\n| 2.0 | N/A | Alpha | N/A | N/A | development ceased |\n\n\nQuestion\nthe download at the top of the chart\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"evolt"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-1303",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-7059.csv` in your code execution environment.\n\nTable Title: Port Adelaide Football Club\nTable Content:\n| Year | Position (after finals) | Coach | Captain | Best & Fairest | Leading goalkicker (goals) |\n| --- | --- | --- | --- | --- | --- |\n| 1997 | 9 | John Cahill | Gavin Wanganeen | Darren Mead | Scott Cummings (70) |\n| 1998 | 10 | John Cahill | Gavin Wanganeen | Adam Kingsley | Warren Tredrea (33) |\n| 1999 | 7 (7) | Mark Williams | Gavin Wanganeen | Stephen Paxman | Warren Tredrea (40) |\n| 2000 | 14 | Mark Williams | Gavin Wanganeen | Brett Montgomery | Warren Tredrea (32) |\n| 2001 | 3 (5) | Mark Williams | Matthew Primus | Warren Tredrea | Warren Tredrea (51) |\n| 2002 | 1 (3) | Mark Williams | Matthew Primus | Matthew Primus | Stuart Dew (51) |\n| 2003 | 1 (4) | Mark Williams | Matthew Primus | Gavin Wanganeen | Warren Tredrea (58) |\n| 2004 | 1 (Premiers) | Mark Williams | Matthew Primus\\nWarren Tredrea1 | Warren Tredrea | Warren Tredrea (81) |\n| 2005 | 8 (6) | Mark Williams | Matthew Primus | Warren Tredrea | Warren Tredrea (65) |\n| 2006 | 12 | Mark Williams | Warren Tredrea | Brendon Lade | Josh Mahoney (29) |\n| 2007 | 2 (Grand Finalist) | Mark Williams | Warren Tredrea | Kane Cornes | Brett Ebert (56) |\n| 2008 | 13 | Mark Williams | Warren Tredrea | Kane Cornes | Daniel Motlop (57) |\n| 2009 | 10 | Mark Williams | Domenic Cassisi | Warren Tredrea | Warren Tredrea (51) |\n| 2010 | 10 | Mark Williams\\nMatthew Primus2 | Domenic Cassisi | Kane Cornes | Jay Schulz (33) |\n| 2011 | 16 | Matthew Primus | Domenic Cassisi | Travis Boak\\nJackson Trengove | Robert Gray (32) |\n| 2012 | 14 | Matthew Primus\\nGarry Hocking3 | Domenic Cassisi | Kane Cornes | Jay Schulz (42) |\n| 2013 | 7 (5) | Ken Hinkley | Travis Boak | Chad Wingard | Jay Schulz (49) |\n| 2014 | | Ken Hinkley | Travis Boak | | |\n\n\nQuestion\nwho had the leading amount of goals kicked in 2004?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"Warren Tredrea"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-7059",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-3229.csv` in your code execution environment.\n\nTable Title: The Boys (comics)\nTable Content:\n| # | Title | TPB ISBN | TPB Release date | TPB page number | Collected material |\n| --- | --- | --- | --- | --- | --- |\n| 1 | The Name of the Game | ISBN 91-33-30546-3 | June, 2007 | 152 | The Boys #1-6 |\n| 2 | Get Some | ISBN 1-933305-68-1 | March, 2008 | 192 | The Boys #7–14 |\n| 3 | Good for the Soul | ISBN 1-933305-92-4 | October, 2008 | 192 | The Boys #15-22 |\n| 4 | We Gotta Go Now | ISBN 1-84856-298-5 | July, 2009 | 192 | The Boys #23-30 |\n| 5 | Herogasm | ISBN 1-60690-082-X | November, 2009 | 144 | Herogasm #1–6 |\n| 6 | The Self-Preservation Society | ISBN 1-60690-125-7 | March, 2010 | 192 | The Boys #31–38 |\n| 7 | The Innocents | ISBN 1-60690-150-8 | December, 2010 | 216 | The Boys #39-47 |\n| 8 | Highland Laddie | ISBN 1-60690-207-5 | April, 2011 | 144 | Highland Laddie #1-6 |\n| 9 | The Big Ride | ISBN 1-60690-220-2 | November 9, 2011 | 276 | The Boys #48-59 |\n| 10 | Butcher, Baker, Candlestickmaker | ISBN 1-60690-264-4 | March 6, 2012 | 144 | Butcher, Baker, Candlestickmaker #1-6 |\n| 11 | Over the Hill with the Sword of a Thousand Men | ISBN 1-60690-341-1 | June/July, 2012 | 152 | The Boys #60-65 |\n| 12 | The Bloody Doors Off | ISBN 1-60690-373-X | December, 2012 | 170 | The Boys #66-72 |\n\n\nQuestion\nwhat was the name of the first title released?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"The Name of the Game"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-3229",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-3891.csv` in your code execution environment.\n\nTable Title: CF Norma San Leonardo\nTable Content:\n| Season | Tier | Division | Place |\n| --- | --- | --- | --- |\n| 1991–98 | 5 | Regional | — |\n| 1998/99 | 4 | 3ª | 8th |\n| 1999/00 | 4 | 3ª | 5th |\n| 2000/01 | 4 | 3ª | 7th |\n| 2001/02 | 4 | 3ª | 9th |\n| 2002/03 | 4 | 3ª | 9th |\n| 2003/04 | 4 | 3ª | 2nd |\n| 2004/05 | 4 | 3ª | 8th |\n| 2005/06 | 4 | 3ª | 6th |\n\n\nQuestion\nthe team finished in the same position in 2001/02 as what other year?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"2002/03"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-3891",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-2075.csv` in your code execution environment.\n\nTable Title: Brock Lesnar\nTable Content:\n| Res. | Record | Opponent | Method | Event | Date | Round | Time | Location | Notes |\n| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |\n| Loss | 5–3 | Alistair Overeem | TKO (kick to the body & punches) | UFC 141 | December 30, 2011 | 1 | 2:26 | Las Vegas, United States | Announced retirement after fight |\n| Loss | 5–2 | Cain Velasquez | TKO (punches) | UFC 121 | October 23, 2010 | 1 | 4:12 | Anaheim, California, United States | Lost UFC Heavyweight Championship |\n| Win | 5–1 | Shane Carwin | Submission (arm triangle choke) | UFC 116 | July 3, 2010 | 2 | 2:19 | Las Vegas, United States | Defended UFC Heavyweight Championship. Unified UFC Interim Heavyweight Championship; Submission of the Night |\n| Win | 4–1 | Frank Mir | TKO (punches) | UFC 100 | July 11, 2009 | 2 | 1:48 | Las Vegas, United States | Defended UFC Heavyweight Championship. Unified UFC Interim Heavyweight Championship |\n| Win | 3–1 | Randy Couture | TKO (punches) | UFC 91 | November 15, 2008 | 2 | 3:07 | Las Vegas, United States | Won UFC Heavyweight Championship. |\n| Win | 2–1 | Heath Herring | Decision (unanimous) | UFC 87 | August 9, 2008 | 3 | 5:00 | Minneapolis, Minnesota, United States | |\n| Loss | 1–1 | Frank Mir | Submission (kneebar) | UFC 81 | February 2, 2008 | 1 | 1:30 | Las Vegas, United States | UFC Debut |\n| Win | 1–0 | Min-Soo Kim | Submission (punches) | Dynamite!! USA | June 2, 2007 | 1 | 1:09 | Los Angeles, United States | MMA Debut |\n\n\nQuestion\nwhat was the last date that lesnar successfully defended the ufc heavyweight belt?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"July 3, 2010"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-2075",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-3997.csv` in your code execution environment.\n\nTable Title: H-II\nTable Content:\n| Flight | Date | Payload nickname | Payload | Orbit | Result |\n| --- | --- | --- | --- | --- | --- |\n| TF1 (Test Flight) | February 4, 1994 | Ryūsei | OREX (Orbital Re-entry Experiment) | LEO | Success |\n| TF1 (Test Flight) | February 4, 1994 | Myōjō | VEP (Vehicle Evaluation Payload) | GTO | Success |\n| TF2 | August 28, 1994 | Kiku 6 | ETS-VI (Engineering Test Satellite-VI) | GEO | Success |\n| TF3 | March 18, 1995 | Himawari 5 | GMS-5 (Geostationary Meteorological Satellite-5) | GEO | Success |\n| TF3 | March 18, 1995 | | SFU (Space Flyer Unit) | LEO | Success |\n| F4 | August 17, 1996 | Midori | ADEOS (Advanced Earth Observing Satellite) | LEO | Success |\n| F4 | August 17, 1996 | Fuji 3 | Fuji OSCAR 29, JAS-2 | LEO | Success |\n| F6 | November 27, 1997 | | TRMM (Tropical Rainfall Measuring Mission) | LEO | Success |\n| F6 | November 27, 1997 | Kiku 7 (Orihime & Hikoboshi) | ETS-VII (Engineering Test Satellite-VII) | LEO | Success |\n| F5 | February 21, 1998 | Kakehashi | COMETS (Communications and Broadcasting Engineering Test Satellites) | GEO | Partial failure |\n| F8 | November 15, 1999 | | MTSAT (Multi-functional Transport Satellite) | GEO | Failure |\n| F7 | Canceled | Midori II | ADEOS-II (Advanced Earth Observing Satellite II) | | Canceled |\n| F7 | Canceled | | μ-LabSat | | Canceled |\n\n\nQuestion\nin what year did the first h-ii flight take place?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"1994"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-3997",
"task_type": "TQA"
}
|
WTQ
|
[
{
"content": "# Mission\nYou are a reasoning agent that solves problems by thinking and using tools in a multi-turn conversation.\n\n# Core Tags\nYou must use the following XML tags to structure your responses:\n- `<think>`: Your private space for reasoning, planning, and evaluating results.\n- `<tool_call>`: Your request to execute a tool. This tag contains the function name and arguments.\n- `<answer>`: Your final, conclusive answer to the user's question, containing only the required JSON object.\n\n# Interaction Protocol\n1. Your interaction is turn-based. Every response you give must start with a `<think>` tag.\n2. After thinking, you can issue one or more `<tool_call>` tags, or present the final `<answer>`.\n3. Example of a `<tool_call>` for table reasoning:\n\n<tool_call>\n{\"name\": \"python\", \"arguments\": {\"code\": \"import pandas as pd\\ndf = pd.read_csv('my_data.csv')\\nprint(df.head())\", \"files\": [\"my_data.csv\"]}}\n</tool_call>\n\nIMPORTANT: Use print() statements to output any results you want to see, otherwise they won't be visible.\n",
"role": "system"
},
{
"content": "# Role\nYou are an expert AI data analyst, skilled at solving complex, table-based questions by combining deep thought with Python code execution.\n\n# Task\nYour mission is to analyze the provided table file and answer a related question. You must achieve this goal through a sequence of thoughts and actions (tool calls).\n\n# Instructions\nPlease follow these steps rigorously:\n\n## 1. Thought\nIn your first response, you must include this section. You need to:\n- Analyze the Problem: Carefully understand the question's requirements and break it down if it is complex.\n- Plan Your Strategy: Design clear steps to solve the problem. Decide which parts require calculation or data extraction via Python code and which parts require subsequent reasoning based on the code's output.\n- Formulate Code (If needed): If you decide to use a tool, formulate the Python code needed to retrieve the key information for your solution.\n\n## 2. Action and Answer\n- Action: After your 'Thought' process, if required, issue a `<tool_call>` to execute your plan.\n- Answer: After you have gathered enough information through thinking and tool calls, you must, in a separate and subsequent response, output only the `<answer>` tag containing the final answer.\n\n# Table File\nYou will have access to a CSV file named `WTQ_nt-1995.csv` in your code execution environment.\n\nTable Title: Canoeing at the 2011 Pan American Games\nTable Content:\n| Rank | Nation | Gold | Silver | Bronze | Total |\n| --- | --- | --- | --- | --- | --- |\n| 1 | Cuba | 4 | 3 | 2 | 9 |\n| 2 | Canada | 4 | 2 | 1 | 7 |\n| 3 | United States | 2 | 0 | 2 | 4 |\n| 4 | Mexico | 1 | 1 | 0 | 2 |\n| 5 | Ecuador | 1 | 0 | 0 | 1 |\n| 6 | Argentina | 0 | 4 | 3 | 7 |\n| 7 | Brazil | 0 | 2 | 2 | 4 |\n| 8 | Chile | 0 | 0 | 1 | 1 |\n| 8 | Venezuela | 0 | 0 | 1 | 1 |\n| Total | Total | 12 | 12 | 12 | 36 |\n\n\nQuestion\nwhich nation won gold but did not win silver?\n\n\nAnswer Format\nThe final answer should be concise and use the following format:\n<answer>\n```json\n{\n \"answer\": [\n \"answer1\",\n \"answer2\",\n ...\n ]\n}\n```\n</answer>\n",
"role": "user"
}
] |
table_reasoning
|
{
"ground_truth": [
"United States"
],
"style": "rule"
}
|
{
"id": "WTQ_nt-1995",
"task_type": "TQA"
}
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.