Datasets:

nawmi
/

evoswe-bench-new

Tasks:

Text Generation

Tags:

License:

Dataset card Files Files and versions

xet

Community

Dataset Viewer

The dataset viewer is not available for this subset.

Cannot get the split names for the config 'default' of the dataset.

Exception:    SplitsNotFoundError
Message:      The split names could not be parsed from the dataset config.
Traceback:    Traceback (most recent call last):
                File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 286, in get_dataset_config_info
                  for split_generator in builder._split_generators(
                                         ~~~~~~~~~~~~~~~~~~~~~~~~~^
                      StreamingDownloadManager(base_path=builder.base_path, download_config=download_config)
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                  )
                  ^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/webdataset/webdataset.py", line 81, in _split_generators
                  first_examples = list(islice(pipeline, self.NUM_EXAMPLES_FOR_FEATURES_INFERENCE))
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/webdataset/webdataset.py", line 32, in _get_pipeline_from_tar
                  fs: fsspec.AbstractFileSystem = fsspec.filesystem("memory")
                                                  ~~~~~~~~~~~~~~~~~^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/fsspec/registry.py", line 302, in filesystem
                  cls = get_filesystem_class(protocol)
                File "/usr/local/lib/python3.14/site-packages/fsspec/registry.py", line 239, in get_filesystem_class
                  raise ValueError(f"Protocol not known: {protocol}")
              ValueError: Protocol not known: memory
              
              The above exception was the direct cause of the following exception:
              
              Traceback (most recent call last):
                File "/src/services/worker/src/worker/job_runners/config/split_names.py", line 71, in compute_split_names_from_streaming_response
                  for split in get_dataset_split_names(
                               ~~~~~~~~~~~~~~~~~~~~~~~^
                      path=dataset,
                      ^^^^^^^^^^^^^
                      config_name=config,
                      ^^^^^^^^^^^^^^^^^^^
                      token=hf_token,
                      ^^^^^^^^^^^^^^^
                  )
                  ^
                File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 340, in get_dataset_split_names
                  info = get_dataset_config_info(
                      path,
                  ...<6 lines>...
                      **config_kwargs,
                  )
                File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 291, in get_dataset_config_info
                  raise SplitsNotFoundError("The split names could not be parsed from the dataset config.") from err
              datasets.inspect.SplitsNotFoundError: The split names could not be parsed from the dataset config.

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.

EvoSWE-Bench (corpus)

Pull requests with their review trajectories — the rounds of reviewer feedback and the commits pushed in response — mined from public GitHub repositories. Built to evaluate coding agents on iterative code-review feedback rather than single-shot issue solving.

Size

TODO -- run round_structure.py and pass --measurements

Rounds: TODO

Structure

file_information.parquet
inline_comment_information.parquet
pr_information.parquet
repository_information.parquet
response_comment_information.parquet
response_commit_information.parquet
rounds_information.parquet
bundles_*.tar.gz — 51 shards of the raw scraped pr.json bundles, ~1,000 PR directories each. Unpack with tar -xzf bundles_0000.tar.gz; each yields one directory per PR containing pr.json and issues/*.json.

The Parquet tables are the primary artifact. They are a relational view of the corpus:

Table	Grain
`repository_information`	one row per repository
`pr_information`	one row per pull request
`rounds_information`	one row per review round
`inline_comment_information`	one row per inline review comment
`response_comment_information`	one row per reply
`response_commit_information`	one row per commit pushed in response
`file_information`	one row per changed file

Things to know before using this

Rounds are clustered, not raw. Use logical_round; round_index counts raw review events, which over-counts roughly 2.5x because one logical round is often several events.

Duplicates are marked, not removed. Filter NOT is_duplicate_round. Whole-round duplication was measured at 1.72%.

files[].patch is not a git diff. It is generated by difflib over two blob texts and carries no diff --git header. Generate patches from a clone; do not git apply this field.

repo_metadata.primary_language is the repository's language, not the PR's. Roughly 16% disagree with the changed files' extensions. Prefer pr_dominant_language.

base_ref_oid is the base branch tip, not the merge base. Compute the merge base per PR if you need the state the PR branched from.

Provenance

Scraped from the GitHub GraphQL API. Several fields carry a *_source column recording whether a value came from the original scrape or a later refetch, so consumers can filter on provenance rather than trusting the weaker value.

Citation

TODO

Downloads last month: 35

Total file size:

20.1 GB