Tag: judge-bias

LLM judges can scale behavioral evaluation, but they inherit biases and can reward style over substance. Reliable use requires clear rubrics, calibration, uncertainty, and human checks.