LLM bias judges name groups instead of saying 'unknown', refusing only 0.4% of the time
ZhijingJin · x · 2026-10-27
A University of Toronto/Vector Institute study at EMNLP 2026 Findings introduces "second-order bias": social bias in how LLM judges evaluate biased content. Asked who would accept a biased text with no demographics given (correct answer: unknown), models name a group anyway — often a dominant one — and refuse only 0.4% of the time. Grounded in entitlement epistemology, the task evades safety guardrails and reveals models' implicit social maps. Code is open-sourced.
Related event: LLMs show second-order bias when judging biased text: EMNLP study(2 posts)→
More from Safety
- Claude's refusal to type one-time passwords is a real security impediment — AaronBergman18 · 2026-10-27
- First systematic data center noise study: 17 of 25 hyperscalers drew no credible complaints — AndyMasley · 2026-10-27
- UK AISI: under $80 of fine-tuning fully restores filtered bio capabilities in open-weight models — connoraxiotes · 2026-10-27
- Meta joins frontier labs reporting CoT controllability evals in Muse Spark 1.3 card — tomekkorbak · 2026-10-27
- Meta TBD underwent pre-launch red-teaming by 5 third-party safety firms — maksym_andr · 2026-10-27
- Is ChatGPT showing your chats to others? Screenshot sparks concern — chrisalbon · 2026-10-27