arXivNataliya Stepanova, Ivan Titov, Emily Allaway, Björn RossTue, Oct 6, 2026, 10:41 AM PDT
score 16.7
New method flags unreliable bias tests in language models
Original: The Missing Minimal Pair: Stereotype Evaluation in LLMs
Source: arxiv.org ↗
Writing ELI5 summary…