新的强化学习框架推动大语言模型推理和效率 · 跟踪 7 个来源

报道来源 [7]

1. arXiv cs.AI TIER_1 English(EN) · Zhicheng Cai, Xinyuan Guo, Hanlin Wu, Weinan Dai, Yuyan Zhou, Wei-Ying Ma, Ya-Qin Zhang, Hao Zhou · 2026-10-12 04:00

   广义近端策略优化：从散度视角统一大语言模型强化学习算法

   arXiv:2610.13015v1 Announce Type: cross Abstract: Reinforcement learning has become the foundational paradigm for enhancing the reasoning capabilities of LLMs. Yet the dominant approach, Proximal Policy Optimization (PPO), relies on a geometrically degenerate clipping proxy that …

2. arXiv cs.AI TIER_1 English(EN) · Leonardo Cesani, Matteo Papini, Marcello Restelli · 2026-10-12 04:00

   方差缩减策略梯度接近最小-最大最优

   arXiv:2610.13692v1 Announce Type: cross Abstract: Standard Policy Gradient (PG) algorithms suffer from high estimation variance because they rely on Monte Carlo updates. While off-policy variance reduction mitigates this problem, establishing optimal sample complexity currently r…

3. arXiv cs.AI TIER_1 English(EN) · Xiang Li, Yuheng Zhang, Nan Jiang · 2026-10-12 04:00

   超越状态-值镜像下降：参数化策略的离线策略优化

   arXiv:2602.23811v5 Announce Type: replace-cross Abstract: We investigate the theoretical aspects of offline reinforcement learning (RL) under general function approximation. While prior works (e.g., Xie et al., 2021) have established the theoretical foundations of learning a good…

4. arXiv cs.LG TIER_1 English(EN) · Du Cheng, Junyu Lu, Jingfan Zhang, Yuzhuo Zhang, Haoxuan Jia, Shichao Weng · 2026-10-12 04:00

   跨行为策略优化：通过共享回归目标统一学习离线演示和在线、离线策略的滚动输出

   arXiv:2610.12830v1 Announce Type: new Abstract: Reasoning post-training draws on offline demonstrations and policy-generated rollouts, yet conventional supervised fine-tuning (SFT) followed by reinforcement learning (RL) separates their objectives and limits continued reuse of co…

5. arXiv cs.LG TIER_1 English(EN) · Zhicheng Cai, Weinan Dai, Yuyan Zhou, Wei-Ying Ma, Ya-Qin Zhang, Hao Zhou · 2026-10-12 04:00

   轨迹感知策略优化：缩小长时域 LLM 强化学习中的状态分布差距

   arXiv:2610.13020v1 Announce Type: new Abstract: Reinforcement learning (RL) has become a dominant paradigm for enhancing LLMs' reasoning capabilities. Yet long-horizon autoregressive rollouts introduce a fundamental limitation: local policy mismatch can compound into severe state…

6. arXiv cs.LG TIER_1 English(EN) · Linxuan Wu, Liang Chen · 2026-10-12 04:00

   面向半精度近端策略优化的结构感知数值稳定

   arXiv:2610.13306v1 Announce Type: new Abstract: Local online training on resource-constrained platforms requires low computational and training-state overhead. We study the numerical stability and computational efficiency of half-precision proximal policy optimization (PPO) on co…

7. arXiv cs.LG TIER_1 English(EN) · El Mahdi Chayti, Alireza Mirrokni, Martin Jaggi · 2026-10-12 04:00

   基于零阶残差跟踪的部分信息随机优化

   arXiv:2610.13871v1 Announce Type: new Abstract: Many optimization problems provide direct access to only part of the gradient, while the remaining component must be inferred from function evaluations. This structure arises in performative prediction, performative reinforcement le…