Reward Modeling

LLM Evaluation

New Rubric Generation Method Improves LLM Judge Accuracy

Researchers propose rethinking how evaluation rubrics are generated for LLM judges and reward models, addressing critical challenges in assessing open-ended AI outputs.