重新思考查询增强中的在线策略优化方法，提出新思路提升检索质量。
arXiv:2510.17139v3 Announce Type: replace Abstract: Recent advances in large language models (LLMs) have led to a surge of interest in query augmentat…
解耦策略学习瓶颈：论文提出RolloutPipe，通过流水线重叠技术并行化LLM强化学习的推理与训练，提升效率。
arXiv:2606.26997v1 Announce Type: cross Abstract: Large language model (LLM) post-training for reasoning increasingly relies on reinforcement learning…
On-Policy蒸馏中策略漂移的块级门控机制，为强化学习知识迁移提供新思路。
arXiv:2606.24084v1 Announce Type: cross Abstract: On-policy distillation (OPD) trains a student policy using teacher signals computed on trajectories …
On-policy蒸馏中旧数据的容忍度如何？这篇论文挑战了传统认知，给出定量分析框架。
arXiv:2606.24143v1 Announce Type: new Abstract: On-policy distillation (OPD) trains a student on its own rollouts guided by teacher feedback and is be…
融合信任区域与在线策略的蒸馏方法，为强化学习知识迁移提供新思路。
arXiv:2606.01249v1 Announce Type: new Abstract: On-Policy Distillation (OPD) is a fundamental technique for efficient post-training of large language …
提出On-Policy Replay方法，解决持续监督微调中的灾难性遗忘问题，为LLM高效增量学习提供新思路。
arXiv:2605.29495v1 Announce Type: new Abstract: Continual supervised fine-tuning (SFT) is the de facto recipe for adapting large language models (LLMs…
最新研究提出VULPO框架，通过on-policy强化学习优化大模型，实现上下文感知的漏洞检测，提升代码安全分析精度。
arXiv:2511.11896v3 Announce Type: replace-cross Abstract: Large language models (LLMs) have recently shown strong potential in vulnerability detection…
蒸馏+低秩适配器让视频生成仅需一两个采样步，颠覆传统扩散流程性能
Distillation + low‑rank tricks cut compute Combining knowledge distillation with low‑rank adapters now yields video generators that need only one or t…
提出f-轨迹平衡损失族，统一了GFlowNets和LLM的on/off-policy训练，梯度对应KL散度，低方差高效。
arXiv:2605.15417v1 Announce Type: cross Abstract: In GFlowNets and variational inference, it has been shown that the mean square error between target …
论文提出on-policy self-distillation方法，在不牺牲推理能力的前提下降低LLM安全对齐中的“安全税”。
arXiv:2605.15239v1 Announce Type: new Abstract: Safety alignment often improves robustness to harmful queries at the cost of reasoning ability, a trad…