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/harbor/harbor.py", line 171, in _split_generators
raise DataFilesNotFoundError("No task.toml or instruction.md files found")
datasets.exceptions.DataFilesNotFoundError: No task.toml or instruction.md files found
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.
WSP-Bench: dev instances
Networks for the WSP-Bench tasks, in which a language model reads, edits or writes neural-network weights. This is
the public dev split: every instance here was built from master seed 42 (instance k of a cell uses the
seed "42.k") and can be rebuilt bit for bit with the generator code (python -m wsb.cli verify) on the same
software and hardware type, both recorded per instance in manifest.json (built_with).
What is included: for each instance, the network as the solver sees it (weights.npz, or several net_*.npz),
its text rendering for no-tool settings (weights_text.txt, when it fits), the network description (subject.md) and
the question (question.md),
and, where the task has them, a tokenizer (tokenizer/) and practice data or vocabularies (materials/).
What is not included: answers, references and scorer data. To score answers on dev, rebuild the instance locally
with the generator (same seed, same result) and use its scorer. Test instances are built with a private salt and are
never published.
Layout: tasks/<task>/D<level>/Z<size tier>/<instance id>/, listed with sha256 of every file in manifest.json.
Z1-Z14 is the size tier (about 3k to 7B parameters, half a decade per tier). D3 is the question version used in the
main table; D1, D2, D4 and D5 are ablation versions (more or less information in the question). Difficulty itself is
set at evaluation time by the solver's tool level: L1 weights only (code, no forward pass), L2 forward passes that
return outputs, L3 outputs plus activations and attention maps, L4 interventions and gradients as well.
| task | name | instances |
|---|---|---|
| A1-1 | GridGame-ID | 24 |
| A1-2 | ImageData-ID | 39 |
| A1-3 | TextTask-ID | 36 |
| A10-1 | Policy-Failure | 1 |
| A10-2 | Injected-Defect | 40 |
| A10-3 | LM-Bug | 1 |
| A11-1 | VLA-Skill-ID | 1 |
| A11-2 | VLA-Grounding-Bug | 1 |
| A11-3 | VLA-Finetune-ID | 1 |
| A2-1 | DDPM-Shapes | 25 |
| A2-2 | Decoder-Factors | 28 |
| A2-3 | CharGen-Format | 1 |
| A2-4 | CharGen-Password | 1 |
| A3-1 | Corpus-ID | 1 |
| A3-2 | Label-Corruption | 40 |
| A3-3 | Augmentation-ID | 28 |
| A4-1 | Optimizer-ID | 40 |
| A4-2 | Checkpoint-Order | 1 |
| A4-3 | Lineage | 1 |
| A5-1 | Head-Roles | 1 |
| A5-2 | Tracr-Program | 1 |
| A5-3 | ModArith | 21 |
| A6-1 | Boolean-Circuit | 25 |
| A6-2 | Real-Function | 25 |
| A6-3 | Automaton-RNN | 25 |
| A6-4 | Rare-Exception | 1 |
| A6-5 | Poison-Substring | 1 |
| A7-1 | Policy-Rank | 1 |
| A7-2 | Classifier-Rank | 41 |
| A7-3 | Weak-Class | 39 |
| A8-1 | Backdoor | 41 |
| A8-2 | Sycophancy-Pair | 1 |
| A8-3 | Hidden-Objective | 1 |
| A9-1 | Classifier-FT-Diff | 39 |
| A9-2 | LM-FT-Diff | 1 |
| A9-3 | Policy-Rule-Change | 1 |
| B1-1 | Kinetix-Repair | 1 |
| B1-2 | GridPolicy-Repair | 1 |
| B1-3 | Classifier-Repair | 40 |
| B1-4 | Hidden-Layer-Repair | 1 |
| B2-1 | Backdoor-Removal | 41 |
| B2-2 | Class-Unlearning | 40 |
| B2-3 | Sycophancy-Removal | 1 |
| B3-1 | Policy-Rule-Insert | 1 |
| B3-2 | Classifier-Remap | 31 |
| B3-3 | Fact-Edit | 1 |
| B4-1 | Loop-Fix | 1 |
| B4-2 | Format-Fix | 1 |
| B4-3 | Over-Refusal-Fix | 1 |
| B5-1 | Arithmetic-Improve | 1 |
| B5-2 | Policy-Improve | 1 |
| B5-3 | Robustness-Improve | 36 |
| B6-1 | Policy-Prune | 1 |
| B6-2 | Classifier-Prune | 34 |
| B6-3 | Head-Prune | 1 |
| B7-1 | VLA-Zone-Rule | 1 |
| B7-2 | VLA-Grounding-Repair | 1 |
| B7-3 | VLA-Prune | 1 |
| B7-4 | VLA-Grounding-Repair-Frozen | 1 |
| C1-1 | GridGame-Write | 25 |
| C1-2 | MinAtar-Write | 1 |
| C1-3 | Go-Write | 1 |
| C2-1 | Boolean-Write | 21 |
| C2-2 | TransformerAlgo-Write | 1 |
| C2-3 | RNN-Automaton-Write | 25 |
| C3-1 | Rule-Tabular-Write | 25 |
| C3-2 | Shape-Image-Write | 30 |
| C3-3 | BoW-Text-Write | 30 |
| C4-1 | Feature-Transplant | 41 |
| C4-2 | Policy-Merge | 1 |
| C4-3 | Induction-Transplant | 1 |
| C5-1 | VLA-Reach-Write | 1 |
| C5-2 | VLA-Skill-Transplant | 1 |
| C5-3 | VLA-Instruction-Write | 1 |
Networks trained on public datasets (images: MNIST, Fashion-MNIST, KMNIST, EMNIST, CIFAR-10, SVHN, Quick, Draw!; texts: SST-2, IMDB, SUBJ, SMS Spam, AG News, TREC, Emotion, Yelp polarity) are released for research; the datasets keep their own licenses. Networks trained on synthetic data or in the benchmark's own grid games are CC BY 4.0.
- Downloads last month
- 329