cs.LG（2025-10-05）

📊 共 7 篇论文

🎯 兴趣领域导航

支柱九：具身大模型 (Embodied Foundation Models) (3) 支柱二：RL算法与架构 (RL & Architecture) (2) 支柱一：机器人控制 (Robot Control) (1) 支柱八：物理动画 (Physics-based Animation) (1)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (3 篇)

#	题目	一句话要点	标签	🔗	⭐
1	DoRAN: Stabilizing Weight-Decomposed Low-Rank Adaptation via Noise Injection and Auxiliary Networks	DoRAN：通过噪声注入和辅助网络稳定权重分解低秩适应	foundation model
2	Beyond Next-Token Prediction: A Performance Characterization of Diffusion versus Autoregressive Language Models	对比扩散模型与自回归语言模型：性能特征分析与优化策略	large language model
3	What Scales in Cross-Entropy Scaling Law?	揭示交叉熵缩放定律失效原因：仅误差熵具有鲁棒缩放性	large language model

🔬 支柱二：RL算法与架构 (RL & Architecture) (2 篇)

#	题目	一句话要点	标签	🔗	⭐
4	Slow-Fast Policy Optimization: Reposition-Before-Update for LLM Reasoning	SFPO：面向LLM推理，通过重定位-更新机制提升强化学习训练效率与稳定性	reinforcement learning large language model
5	Simple Policy Gradients for Reasoning with Diffusion Language Models	提出AGRPO算法，用于扩散语言模型推理的策略梯度优化	reinforcement learning large language model

🔬 支柱一：机器人控制 (Robot Control) (1 篇)

#	题目	一句话要点	标签	🔗	⭐
6	A KL-regularization framework for learning to plan with adaptive priors	提出PO-MPC框架，通过KL正则化学习自适应先验的规划策略，提升MBRL性能。	MPC model predictive control reinforcement learning

🔬 支柱八：物理动画 (Physics-based Animation) (1 篇)

#	题目	一句话要点	标签	🔗	⭐
7	Wave-PDE Nets: Trainable Wave-Equation Layers as an Alternative to Attention	提出Wave-PDE Nets，以可训练波动方程层替代注意力机制，提升计算效率。	differentiable simulation

⬅️ 返回 cs.LG 首页 · 🏠 返回主页