Awesome RL with Human Feedback / 2025 / arXiv

A Unified Pairwise Framework for RLHF: Bridging Generative Reward Modeling and Policy Optimization

作者：Wenyuan Xu, Xiaochen Zuo, Chao Xin, Yu Yue, Lin Yan, Yonghui Wu · 单位：ByteDance Seed · 来源：arXiv · 发布日期：2025

奖励建模偏好学习策略优化框架/系统
新标签打开 AlphaXiv
A Unified Pairwise Framework for RLHF: Bridging Generative Reward Modeling and Policy Optimization figure
AlphaXiv 中文概览（可滚动查看）