Awesome RL with Human Feedback / 2023 / arXiv

Aligning Language Models with Preferences through f-divergence Minimization

作者：Dongyoung Go, Tomasz Korbak, Germán Kruszewski, Jos Rozen, Nahyeon Ryu, Marc Dymetman · 来源：arXiv · 发布日期：2023

偏好学习对齐目标设计训练稳定性策略优化策略梯度人类反馈强化学习
新标签打开 AlphaXiv
Aligning Language Models with Preferences through f-divergence Minimization figure
AlphaXiv 中文概览（可滚动查看）