**New on Benchmark Radar: FairStress** (Allen Institute for AI) A deeper evalu… · @ktwu01 · TwKit

@ktwu01

**New on Benchmark Radar: FairStress** (Allen Institute for AI) A deeper evaluation of fairness in LLM decision-making. Models that look unbiased on standard tests show real disparities once decisions come with time pressure, missing information, and competing priorities. FairStress also catches demographic gaps in how models justify their choices, even when the final choices look fair. 300 decision scenarios across 29 high-stakes domains, 12,000+ cases. Now indexed on Benchmark Radar. Announcement: https://huggingface.co/blog/allenai/fairstress Radar: https://benchmark-radar.org

曝光 20 · 评论 0 · 点赞 0 · 书签 0