In-Situ Behavioral Evaluation for LLM Fairness, Not Standardized-Test Scores

Zeyu Tang, Sang T. Truong, Deonna Owens +4

LLM fairness should be evaluated through in-situ behavioral pattern rather than standardized-test Q&A benchmarks. We show that the standardized-test paradigm can be structurally un…