1 paper · 1 filter
Daein Weon, Dong Ho Kang
Evaluation scores used around LLM systems -- including reward models, rerankers, and LLM judges -- can track surface form instead of the quality they claim to measure. When present…