catt-bench/vtok101-distr-attribution-baselines

Permissive or share-alike

The declared licence (mit) is classified as permissive.

Licence facts reported by the dataset and its declared sources. Not legal advice, and not a grant of permission to use the dataset.

Evidence

• Declared licence: mit (from the tag)
• Licence files: no licence file at the top level of the repository
• Access gated on Hugging Face: no

Declared sources

No source datasets listed in the card metadata.

Signal

See all datasets marked Permissive or share-alike.

Citation text from the card

The attribution scores of "Elucidating the Design Space of LM Data Attribution"
on **Function Bindings with distractors**, for the LoRA adapters in
[catt-bench/Qwen3.5-4B-d0-vtok101-distr-lora-seeds](https://huggingface.co/catt-bench/Qwen3.5-4B-d0-vtok101-distr-lora-seeds). The CATT code release reads them
with `catt fetch results --task fb-distr` and rebuilds the paper's figures and tables
from them with `catt report results`.

Every real document has a decoy beside it: a shadow function `<Axx>` taught the
same constant in the same template as `<Bxx>`. The pool is twice as large and
queries only ask about real functions, so AUC keeps its meaning while Recall@N's
chance line halves. `metrics.json` adds `decoy_auc` (a real document against the
query's own decoys) and `decoy_top1`.

Downloads (30 days): 9869. Likes: 4. Last modified: 2026-10-29T18:11:38.000Z.

View catt-bench/vtok101-distr-attribution-baselines on Hugging Face | JSON