A 328-item BIG-bench free-response task that names the shared category of a short list of words or objects.
unassessed
| Category | knowledge |
|---|---|
| Subcategory | BIG-bench free-response hypernym / category naming (328 items) |
| Page status | unknown |
| Metric | bleu |
| Direction | higher_is_better |
| Dataset size | 328 |
| Dataset licence | Apache-2.0 |
| Publisher | Google (BIG-bench collaboration) |
auto_categorization gives a model a short English list of words, names or titles and asks it to name the broader class they belong to. Author Fanyue Xia built the items from Kaggle and Wikidata properties, grouping up to three examples per property under a varied sentence template. The intended skill is common-sense taxonomy, not a closed label set: the target is a word or phrase such as "colors" or "pets", and 95 items list more than one acceptable phrasing.
Free-response JSON task scored with BIG-bench BLEU. A canary GUID is embedded. There is no multiple-choice variant in the task directory.
No model card in ModelSpec reports this benchmark yet.