作者位次 #4/9ACL 2026Findings· Findings of ACL
CE - RM : A Pointwise Generative Reward Model Optimized via Two-Stage Rollout and Unified Criteria
Xinyu Hu · Yancheng He · Weixun Wang · Tao Feng · Li Lin · Jiashun Liu · Wenbo Su · Zerui Huang (Taobao & · Xiaojun Wan
摘要
Automatic evaluation is crucial yet challenging for open-ended natural language generation, especially when rule-based metrics are infeasible. Compared with traditional methods, the recent LLM-as-a-Judge paradigms enable better and more flexible evaluation, and show promise as generative reward m...