Awesome RL with Human Feedback / 2025 / arXiv

Differential Information: An Information-Theoretic Perspective on Preference Optimization

作者：Yunjae Won, Hyunji Lee, Hyeonbin Hwang, Minjoon Seo · 来源：arXiv · 发布日期：2025

奖励建模偏好学习策略优化数据筛选理论/遗憾界
新标签打开 AlphaXiv
Differential Information: An Information-Theoretic Perspective on Preference Optimization figure
AlphaXiv 中文概览（可滚动查看）