Awesome RL with Human Feedback / 2026 / arXiv

Causally Robust Reward Learning from Reason-Augmented Preference Feedback

作者：Minjune Hwang, Yigit Korkmaz, Daniel Seita, Erdem Biyik · 单位：Thomas Lord Department of Computer Science, University of Southern California · 来源：arXiv · 发布日期：2026

鲁棒性/噪声反馈奖励建模偏好学习
新标签打开 AlphaXiv
Causally Robust Reward Learning from Reason-Augmented Preference Feedback figure
AlphaXiv 中文概览（可滚动查看）