Aznaur/terminal_agent_tbench_pro_v2_split

Terminal Agent - Multi-Task NAT v13

Model Description

This model is fine-tuned from Qwen3-8B on multi-task terminal agent trajectories using Negative-Aware Training (NAT).

Key Features

• 5 Tasks: fix-git, cancel-async-tasks, log-summary-date-ranges, regex-log, pypi-server
• Fixed Tool Signatures: Corrected critical bug where note_name was incorrectly removed
• Clean Tool Calls: Removed hallucinated parameters (message_title, message_description, message_attachment)
• Negative Examples: Includes looping and wrong_command negative examples

Training Details

• Base Model: Qwen/Qwen3-8B
• Training Data: 40 samples (20 positive, 20 negative)
• Epochs: 300
• Learning Rate: 5e-5
• Batch Size: 4

Tool Signatures (Corrected)

• shell_exec(id, command, block)
• shell_write_content_to_file(content, file_path)
• create_note(note_name, content)
• append_note(note_name, content)
• read_note(note_name)

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("camel-ai/terminal_agent_multitask_nat_v13")
tokenizer = AutoTokenizer.from_pretrained("camel-ai/terminal_agent_multitask_nat_v13")

V13 Fixes

1. KEEP note_name - Required by runtime (was incorrectly removed in v12)
2. System prompt uses note_name - Matches runtime expectations
3. Remove only hallucinated params - message_title, message_description, message_attachment
4. Added tool call validation - Catches signature issues before training

Evaluation Results

Expected to achieve >80% success rate on 5 tasks when evaluated with matching task set.

License

MIT License