Dataset Viewer

Split (1)

train

The dataset viewer is not available for this split.

Cannot load the dataset split (in streaming mode) to extract the first rows.

Error code:   StreamingRowsError
Exception:    CastError
Message:      Couldn't cast
instance_id: string
repo: string
base_commit: string
patch: string
test_patch: string
problem_statement: string
hints_text: string
created_at: timestamp[s]
version: string
FAIL_TO_PASS: list<item: string>
  child 0, item: string
PASS_TO_PASS: list<item: string>
  child 0, item: string
dockerfile: string
environment: struct<base_image: string, node_version: string, package_manager: string, install_commands: list<ite (... 33 chars omitted)
  child 0, base_image: string
  child 1, node_version: string
  child 2, package_manager: string
  child 3, install_commands: list<item: string>
      child 0, item: string
  child 4, test_command: string
pins: struct<base_image: string, n: string, cached_node_versions: string, yarn: string, pnpm: string, bun: (... 8 chars omitted)
  child 0, base_image: string
  child 1, n: string
  child 2, cached_node_versions: string
  child 3, yarn: string
  child 4, pnpm: string
  child 5, bun: string
build_host: string
image_size_bytes: int64
built_at: timestamp[s]
image: string
apt_mirror_security: string
image_id: string
image_digest: string
apt_mirror: string
build_note: string
build_arch: string
to
{'image': Value('string'), 'image_id': Value('string'), 'image_digest': Value('string'), 'image_size_bytes': Value('int64'), 'built_at': Value('timestamp[s]'), 'build_host': Value('string'), 'build_arch': Value('string'), 'dockerfile': {'path': Value('string'), 'sha256': Value('string'), 'git_commit': Value('string')}, 'pins': {'base_image': Value('string'), 'n': Value('string'), 'cached_node_versions': Value('string'), 'yarn': Value('string'), 'pnpm': Value('string'), 'bun': Value('string')}, 'apt_mirror': Value('string'), 'apt_mirror_security': Value('string'), 'build_note': Value('string')}
because column names don't match
Traceback:    Traceback (most recent call last):
                File "/src/services/worker/src/worker/utils.py", line 147, in get_rows_or_raise
                  return get_rows(
                      dataset=dataset,
                  ...<4 lines>...
                      column_names=column_names,
                  )
                File "/src/libs/libcommon/src/libcommon/utils.py", line 272, in decorator
                  return func(*args, **kwargs)
                File "/src/services/worker/src/worker/utils.py", line 127, in get_rows
                  rows_plus_one = list(itertools.islice(safe_iter(ds, dataset=dataset), rows_max_number + 1))
                File "/src/services/worker/src/worker/utils.py", line 483, in safe_iter
                  yield from ds.decode(False) if ds.features else ds
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2840, in __iter__
                  for key, example in ex_iterable:
                                      ^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2373, in __iter__
                  for key, pa_table in self._iter_arrow():
                                       ~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2398, in _iter_arrow
                  for key, pa_table in self.ex_iterable._iter_arrow():
                                       ~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 536, in _iter_arrow
                  for key, pa_table in iterator:
                                       ^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 419, in _iter_arrow
                  for key, pa_table in self.generate_tables_fn(**gen_kwags):
                                       ~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 343, in _generate_tables
                  self._cast_table(pa_table, json_field_paths=json_field_paths),
                  ~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 132, in _cast_table
                  pa_table = table_cast(pa_table, self.info.features.arrow_schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2378, in table_cast
                  return cast_table_to_schema(table, schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2306, in cast_table_to_schema
                  raise CastError(
                  ...<3 lines>...
                  )
              datasets.table.CastError: Couldn't cast
              instance_id: string
              repo: string
              base_commit: string
              patch: string
              test_patch: string
              problem_statement: string
              hints_text: string
              created_at: timestamp[s]
              version: string
              FAIL_TO_PASS: list<item: string>
                child 0, item: string
              PASS_TO_PASS: list<item: string>
                child 0, item: string
              dockerfile: string
              environment: struct<base_image: string, node_version: string, package_manager: string, install_commands: list<ite (... 33 chars omitted)
                child 0, base_image: string
                child 1, node_version: string
                child 2, package_manager: string
                child 3, install_commands: list<item: string>
                    child 0, item: string
                child 4, test_command: string
              pins: struct<base_image: string, n: string, cached_node_versions: string, yarn: string, pnpm: string, bun: (... 8 chars omitted)
                child 0, base_image: string
                child 1, n: string
                child 2, cached_node_versions: string
                child 3, yarn: string
                child 4, pnpm: string
                child 5, bun: string
              build_host: string
              image_size_bytes: int64
              built_at: timestamp[s]
              image: string
              apt_mirror_security: string
              image_id: string
              image_digest: string
              apt_mirror: string
              build_note: string
              build_arch: string
              to
              {'image': Value('string'), 'image_id': Value('string'), 'image_digest': Value('string'), 'image_size_bytes': Value('int64'), 'built_at': Value('timestamp[s]'), 'build_host': Value('string'), 'build_arch': Value('string'), 'dockerfile': {'path': Value('string'), 'sha256': Value('string'), 'git_commit': Value('string')}, 'pins': {'base_image': Value('string'), 'n': Value('string'), 'cached_node_versions': Value('string'), 'yarn': Value('string'), 'pnpm': Value('string'), 'bun': Value('string')}, 'apt_mirror': Value('string'), 'apt_mirror_security': Value('string'), 'build_note': Value('string')}
              because column names don't match

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

TS SWE-bench 数据集交付包

松果基金 SWE 数据验收问题修复后的最终交付。包含：1013 条 TS SWE-bench 数据集、基础镜像构建资产（写死版本 Dockerfile + 构建脚本 + 构建清单）、评测消费脚本与全套文档。

目录结构

delivery/
├── data/
│   └── swebench-2026-08-19T00-00-00-fixed-b2-b4-b1-b3-b4f-node-b2fix-b4d-supp-1013.jsonl   # 1013 条数据集
├── docker/
│   ├── Dockerfile.ts-bench-base        # 基础镜像（写死版本：base digest + n/yarn/pnpm/bun 固定 + 8 个 Node 主版本预缓存 + 37 个 apt 包 + corepack yarn4 缓存）
│   ├── build-base.sh                   # 本地构建脚本（输出 MANIFEST）
│   └── MANIFEST-2026-08-19.json        # 构建清单（image_id / digest / pins / 溯源）
├── evaluator/
│   └── consume-swebench.js             # 评测消费脚本（含 Node 版本自动切换）
├── README.md                           # 本文件
├── 修复报告.md                          # 验收问题 → 修复措施 → 结果（逐项）
├── 数据说明.md                          # 数据集 schema / 统计 / 质量校验
└── 留底说明.md                          # base_image_digest 与镜像 digest 溯源

快速使用

1. 构建基础镜像（消费方本地构建，不依赖推送）

cd docker
bash build-base.sh 2026-08-19
# 可选：使用国内 apt 镜像加速
# APT_MIRROR=http://mirrors.aliyun.com/debian \
# APT_MIRROR_SECURITY=http://mirrors.aliyun.com/debian-security \
#   bash build-base.sh 2026-08-19

构建产物：本地镜像 ts-bench-base:2026-08-19（digest 见 MANIFEST-2026-08-19.json，可核对）。

2. 评测

方式 A：标准 SWE-bench harness 每条实例的 dockerfile 字段为 FROM ts-bench-base:2026-08-19 + RUN n <主版本>，直接使用标准 harness 的 build_image / run_evaluation 即可（Node 版本由 dockerfile 离线切换）。

方式 B：自带评测脚本

cd evaluator
node consume-swebench.js \
  --input ../data/swebench-2026-08-19T00-00-00-fixed-b2-b4-b1-b3-b4f-node-b2fix-b4d-supp-1013.jsonl \
  --output-dir ./results \
  --workers 4 \
  --repo-cache-dir ./repo-cache

说明：该脚本按 environment.node_version 主版本自动执行 n <主版本> 切换（离线、瞬时），无需每实例单独构建镜像。

3. 数据格式

标准 SWE-bench 格式，每条实例 13 个顶层字段： instance_id / repo / base_commit / problem_statement / hints_text / patch / test_patch / FAIL_TO_PASS / PASS_TO_PASS / version / created_at / dockerfile / environment

其中 environment 为标准 5 字段：base_image / install_commands / node_version / package_manager / test_command。

关键事实速览

数据集：1013 条 / 312 仓库
F2P 用例：4,480 个（含重复，唯一 4,406）
基础镜像：ts-bench-base:2026-08-19（digest sha256:d1ab53e7...，见 MANIFEST）
Node 版本：默认 20.20.2，预缓存 8/14/16/18/20/22/23/24 全部主版本，离线切换
修复状态：B1–B5 全部为 0（详见 修复报告.md）

Downloads last month: 45

Total file size:

19.8 MB