Feedback noise is pervasive in preference data, arising from subjective human judgments, annotation mistakes, and imperfect LLM judges. Standard reward model…
机构:浙大
来源:arXiv 2610.21138 | AI4Papers 论文推荐平台