arXiv 2610.08963 · cs.LG · 2026-10-06
On KL-Regularized Policy Optimization
Yifan Zhang
math-ai
在最新榜 #228 本站收录于 2026-10-21
一句话要点
KLPO 把 KL 正则锚定在采样策略上，处理异步强化学习中策略过期和推理/训练概率不一致的问题。
KLPO anchors the KL regulariser at the sampling policy to handle stale, mismatched rollouts in asynchronous LLM-agent RL.
阅读 arXiv 原文 PDF Hugging Face 讨论 代码仓库 项目页
主题与学科
同主题：强化学习
- Waypoint-OPD: Transferring RL-Induced Policy Shifts via Gated On-Policy Distillation ▲219
- Self-Retrospection Distillation: Turning Post-hoc Experiences into Prior Foresight ▲189
- LoGRA: Scaling LLM Reinforcement Learning with Low-Rank Gradient Sketches ▲152
- Sharpening Tax in Post-Training ▲111
- RULER: Instance-aware Rubric Rewards for SVG Generation ▲104
数据说明
元数据来自 arXiv API（已核实）；点赞来自 Hugging Face 每日论文；引用来自 Semantic Scholar。最近刷新 2026-10-23。
要点由本站根据摘要自写，不转载摘要原文，请阅读原文。