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.
NeurIPS 2026 GenericAgent traces
Exactly 300 distinct Test tasks per model, across seven model cohorts (2,100 runs). Each archive contains one selected trace per task with a valid benchmark outcome.
Contents
| Model cohort | Valid runs | Files | Archive GB |
|---|---|---|---|
| qwen3.5-9b | 300 | 24,581 | 5.70 |
| qwen3.8-27b | 300 | 25,574 | 6.39 |
| gemma-4-26b-a4b-it | 300 | 23,367 | 6.08 |
| gemma-4-31b-it | 300 | 21,814 | 6.06 |
| gemini-3.1-flash-lite-preview | 300 | 21,566 | 5.16 |
| gemini-3.1-pro-preview-omni | 300 | 23,084 | 6.60 |
| claude-opus-4.7 | 300 | 25,364 | 6.96 |
Each test/<model>/task-<id>/ directory contains:
episode/: the selected original trace, screenshots, step objects, and arguments.outcome/summary_info.jsonandoutcome/task_info.json: the authoritative selected outcome.selection.json: portable paths, source provenance, and hashes.provenance/andevidence/, where applicable: recorded regrades and native evidence.
Original episode bytes are preserved. For repaired or regraded runs, the original
episode summary can differ from the selected outcome. Always read outcome/ or use
the loader below. The per-model metadata/*.jsonl indices contain exactly 300 rows.
file_inventory.jsonl.gz records every archive member and its SHA-256.
This is a file archive dataset; it is not a normalized Hugging Face Datasets table.
Download and load
hf auth login
hf download WebArena-Pro/neurips_2026_traces_generic_agent --repo-type dataset --local-dir neurips_2026_traces_generic_agent
cd neurips_2026_traces_generic_agent
shasum -a 256 -c SHA256SUMS
mkdir -p extracted
for archive in archives/test/*.tar; do
tar -xf "$archive" -C extracted
done
python load_traces.py --model qwen3.5-9b
from load_traces import load_traces
traces = list(load_traces(".", "extracted", "qwen3.5-9b"))
assert len(traces) == 300
for trace in traces:
print(trace["task_id"], trace["reward"], trace["trace_dir"])
Download a single model archive:
from huggingface_hub import hf_hub_download
path = hf_hub_download(
repo_id="WebArena-Pro/neurips_2026_traces_generic_agent",
repo_type="dataset",
filename="archives/test/gemini-3.1-pro-preview-omni.tar",
token=True,
)
- Downloads last month
- 7