CrowS-Pairs

Crowdsourced English sentence pairs that score whether a language model assigns higher likelihood to the more stereotyping sentence than to its minimally edited counterpart.

Also known as: Crowdsourced Stereotype Pairs, crows-pairs, crows_pairs_english

unassessed

This page is a discovery lead. Nobody has yet assessed it against the catalogue contract, so it carries no disposition. Absence of evidence here is not evidence of staleness.
Categorysafety
SubcategoryUS English stereotype-pair likelihood comparison across nine bias types
Page statusunknown
Metricpct_stereotype (share of pairs where the more-stereotyping sentence is more likely)
Directionlower_is_better
Unit%
Dataset size1508
Dataset licenceCC BY-SA 4.0
PublisherNew York University, Machine Learning for Language group

What it measures

CrowS-Pairs (Crowdsourced Stereotype Pairs) tests whether a language model prefers US social stereotypes. Each item is two English sentences that differ only in the words that name a historically disadvantaged group versus a contrasting advantaged group. One sentence is more stereotyping (sent_more); the other is less stereotyping (sent_less). The nine types are race/color, gender/gender identity, socioeconomic status/occupation, nationality, religion, age, sexual orientation, physical appearance, and disability. The original metric was built for masked language models: compare conditional pseudo-log-likelihoods of the two sentences without masking the modified tokens. lm-evaluation-harness reuses the same pair test for causal models by comparing full-sentence log-likelihoods.

Task format

Paired likelihood comparison, no generated answer. lm-eval sets output_type multiple_choice with empty doc_to_text, choices sent_more and sent_less, and target index 0 (sent_more). Zero-shot. The original metric.py supports bert, roberta, and albert only.

Models reporting this benchmark

No model card in ModelSpec reports this benchmark yet.

Data

This page as JSON · Edit on GitHub