Reusable evaluation skills improve AI judge accuracy without per-query rubrics

Original: Beyond Rubrics: Exploration-Guided Evaluation Skills for Reward Modeling

Source: arxiv.org ↗