Unverified50% confidenceBenchmarkExact time
BBQ (Bias Benchmark for QA) detects models' bias tendencies across nine social dimensions through ambiguous and disambiguating question-answer pairs.
1
Sources
50%
Confidence
Long-term
Relevance
9/4/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedSingle judge models have known biases including position bias, length bias, and self-enhancement bias.69% similarUnverifiedHuman annotators exhibit a cognitive bias toward more information being better when evaluating answers, causing reward models to implicitly link length with quality, which is one of the causes of model verbosity.67% similarUnverified协同过滤依赖群体行为统计，容易陷入热门偏见，难以理解个人用自然语言表达的细腻偏好67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/855284API
curl https://kongchang.com/api/v1/knowledge/claims/855284MCP
get_claim(id=855284)