Unverified50% confidenceFactExact time
BBQ is a dataset specifically designed to evaluate social bias in question-answering models, covering multiple dimensions such as gender, race and age, with questions divided into ambiguous-context and unambiguous-context categories.
1
Sources
50%
Confidence
Long-term
Relevance
10/30/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedBBQ (Bias Benchmark for QA) detects models' bias tendencies across nine social dimensions through ambiguous and disambiguating question-answer pairs.75% similarUnverifiedMainstream bias benchmarks currently include datasets such as WinoBias, StereoSet, and BBQ, which measure a model's stereotyping tendencies along dimensions such as gender and race by constructing contrastive sentence pairs or multiple-choice questions.66% similarUnverifiedThis fairness audit framework centers on paired A/B scenarios plus symmetry scoring, with each scenario changing only one demographic parameter (name, gender, class, place of origin, city, or political sensitivity) while all other conditions are held fixed.58% similarUnverified位置偏差是偏好数据集中一个被广泛研究的系统性问题，人类标注者评估并排回答时会受展示顺序影响55% similarUnverifiedSBIC+ (an extended version of the Social Bias Inference Corpus) is a broader social bias dataset covering multiple bias types.55% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1115480API
curl https://kongchang.com/api/v1/knowledge/claims/1115480MCP
get_claim(id=1115480)