kensho/WILD
Dataset to accompany the paper Cost-Efficient Estimation of General Abilities Across Benchmarks. WILD: Wide-scale Item Level Dataset WILD is a large-scale evaluation response matrix containing item-level binary scores for 65 language models across 27 benchmarks (163 subtasks, 109,566 unique items). Each row records whether a model answered a specific evaluation item correctly, along with token usage. For full response content (conversations, targets, model answers, scorer… See the full description on the dataset page: https://huggingface.co/datasets/kensho/WILD.
| Author / organization | kensho |
|---|---|
| License (as declared) | apache-2.0 |
| Downloads | 263 (as reported by Hugging Face) |
| Likes | 2 |
| Created | 2 April 2026 |
| Last updated at source | 7 May 2026 |
Popularity, last 2 days
263 now (+45 since 2026-10-16). Counts as reported by the source, snapshotted daily by Silytics.
Connected in the knowledge graph
Recent changes
- 16 Oct 2026 Description: Dataset to accompany the paper Cost-Efficient Estimation of General Abilities Across Benchmarks. WILD: Wide-scale Item Level Dataset WILD is a large-scale evaluation response matrix containing item-level binary scores for 65 language models → Dataset to accompany the paper Cost-Efficient Estimation of General Abilities Across Benchmarks. WILD: Wide-scale Item Level Dataset WILD is a large-scale evaluation response matrix containing item-level binary scores for 65 language models
Changes detected by Silytics between refreshes of the source.
Related papers
More from kensho
- kensho/DocFinQADataset
- kensho/PubTables-v2Dataset
- kensho/SPGISpeech2.0Dataset
- kensho/spgispeechDataset
- kensho/bizbenchDataset
- kensho/WILD-rawDataset
Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 17 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.