从EM视角透视RL训练推理模型，揭示PPO/GRPO的数学本质，为优化大模型推理提供新思路。
arXiv:2504.18587v2 Announce Type: replace-cross Abstract: Reinforcement learning has emerged as a powerful approach for improving the reasoning capabi…
无需投影的线性TD(0)用单一固定步长即可同时获得稳健与快速收敛，理论突破值得关注。
arXiv:2606.24981v1 Announce Type: new Abstract: We study linear TD(0) under Markovian sampling, where data are generated along a single trajectory. We…
探索测度近似的新结构化方法，为机器学习和概率论提供理论工具
arXiv:2310.09149v3 Announce Type: replace-cross Abstract: We study the approximation of probability measures in the Wasserstein-$p$ distance by struct…
用公平赌局的平均结果重新定义评估指标，揭示评测背后的数学本质，适合关注模型评估理论的研究者。
arXiv:2401.14483v4 Announce Type: replace Abstract: In the current practices of machine learning, the evaluation of forecasts has become a cornerstone…
从数学底层揭示多种相位恢复算法在特定条件下等价，突破传统认知。
arXiv:2606.07257v1 Announce Type: cross Abstract: Phase retrieval - recovering a complex-valued field from intensity measurements - is typically solve…
从变分角度揭示Föllmer过程在生成扩散中的最优性，理论数学与AI生成的深度交叉
arXiv:2602.10989v2 Announce Type: replace-cross Abstract: We construct and analyze generative diffusions that transport a point mass to a prescribed t…