Datasets:
The dataset viewer is not available for this subset.
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 84, in _split_generators
raise ValueError(
...<2 lines>...
)
ValueError: The TAR archives of the dataset should be in WebDataset format, but the files in the archive don't share the same prefix or the same types.
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 68, 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.
WorkBuddy Bench — Datasets
The task datasets for WorkBuddy Bench, a benchmark for evaluating coding agents on real-world developer, PM, algo, QA, ops, and security work. This repository hosts only the task data; the evaluation framework, setup, and usage all live in the GitHub repository above.
Dataset Summary
WorkBuddy Bench is built from real-world work tasks. Given a task and a
sandboxed workspace, an agent must produce the correct change — a patch, an
artifact, or a report — which is then graded against a test suite. The benchmark
ships four independent subsets, one .tar.gz archive each:
| Subset | Archive | Tasks | Domain |
|---|---|---|---|
| Code | wb-bench-code-v1.0.tar.gz |
80 | repo-level coding tasks |
| Web | wb-bench-web-v1.0.tar.gz |
70 | front-end / GUI (HTML/CSS/JS) |
| Office | wb-bench-office-v1.0.tar.gz |
50 | office data / file workflows |
| Security | wb-bench-sec-v1.0.tar.gz |
60 | security / vuln (red + blue) |
SHA256SUMS lists the checksum of every archive.
Dataset Structure
Each archive extracts to wb-bench-<subset>-v1.0/, with one directory per task:
wb-bench-<subset>-v1.0/
dataset.toml
tasks/<task-id>/
instruction.md # the task prompt
task.toml # task metadata
environment/ # Dockerfile, docker-compose, scorer, workspace.tar.gz
tests/ # grading suite
Each task's environment/workspace.tar.gz stays packed; the task Dockerfile
extracts it into /workspace inside the container.
Usage
These archives are consumed by the WorkBuddy Bench framework, not loaded directly. See the GitHub repository for how to download, extract, and run evaluations.
- Downloads last month
- 68