cs.LG（2026-08-26）

📊 共 25 篇论文 | 🔗 2 篇有代码

🎯 兴趣领域导航

支柱九：具身大模型 (Embodied Foundation Models) (12 🔗2) 支柱二：RL算法与架构 (RL & Architecture) (7) 支柱八：物理动画 (Physics-based Animation) (3) 支柱一：机器人控制 (Robot Control) (1) 支柱七：动作重定向 (Motion Retargeting) (1) 支柱四：生成式动作 (Generative Motion) (1)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (12 篇)

#题目一句话要点标签🔗⭐
1 A General-Purpose Molecular Foundation Model Transfers Across Diverse Olfactory Tasks 提出通用分子基础模型以解决多样化嗅觉任务 foundation model
2 Large Language Model Few-Shot Prompting with Dilemma Training Outperforms Human Surrogates in Predicting Patient Preferences 提出P4-DT以解决患者偏好预测的准确性问题 large language model
3 Why Does Graph Learning Fail to Fully Benefit from a Text Teacher? 提出多模态模型以解决图学习与文本教师结合的挑战 large language model multimodal
4 TraceML: An Empirical Analysis of Human-Agent Planning in Machine Learning Development 提出TraceML以分析人类与智能体在机器学习开发中的规划差异 large language model ✅
5 Spectral Allocation: Why Muon Outperforms Adam, and How to Improve Muon 提出Spectral-Aware Muon以提升大语言模型训练效率 large language model
6 When Pruning Meets Interpretability: Preserving Sparse Autoencoder Robustness in LLMs 提出层级稀疏分配策略以提升稀疏自编码器在大语言模型中的鲁棒性 large language model ✅
7 EXAONE Tabular 1.0 : Technical Report 提出EXAONE Tabular以提升表格数据的分类与回归性能 foundation model
8 Are LLM-Enhanced GNNs Privacy-Safe? 系统评估LLM增强GNN的隐私风险与防护策略 large language model
9 Beyond Scaling: Self-Evolving LLM Agents for Hardware Kernel Optimization via an Experience-Driven Workflow and Experience Graph Memory 提出KOPE框架以优化硬件内核的自动化过程 foundation model
10 Escaping Low-Dimensional Overlap: Multi-Task Model Merging via High-Dimensional Sparse Disentanglement 提出稀疏表示合并框架以解决多任务模型重叠问题 instruction following
11 InsightSR: Refining Symbolic Regression Search Spaces via Parallel Semantic and Structural LLM Guidance 提出InsightSR以优化符号回归搜索空间 large language model
12 Mitigating LLM sycophancy with RL-based fine-tuning: Bayesian Truth Serum approach 提出基于RL的细化方法以缓解LLM的谄媚现象 large language model

🔬 支柱二：RL算法与架构 (RL & Architecture) (7 篇)

#题目一句话要点标签🔗⭐
13 Drift-Aware Multimodal User Representation Learning via Multi-Scale Temporal Modeling and Sparse Mixture-of-Experts 提出DUMoE框架以解决社交媒体用户偏好漂移问题 representation learning multimodal
14 One Symptom, Three Levers: A Critical Review of On-Policy Self-Distillation 提出自蒸馏方法以解决模型推理路径收敛问题 reinforcement learning imitation learning distillation
15 Cooperative Multi-Agent Reinforcement Learning for Adaptive Aggregation in Semi-Supervised Federated Learning with non-IID Data 提出pFedMARL以解决非IID数据下的联邦学习聚合问题 reinforcement learning TD3
16 DualOPSD: Adaptive Privileged Teachers for On-Policy Self-Distillation 提出DualOPSD以解决OPSD固定教师适应性不足问题 distillation
17 A Token-Level Analysis of Sampled-Token Reverse-KL On-Policy Distillation 提出SuRe以优化样本令牌反向KL的在线蒸馏过程 distillation
18 BVR Sim: An Open and High-Throughput Environment for Heterogeneous Air-Combat Reinforcement Learning 提出BVR Sim以解决异构空战强化学习环境不足问题 reinforcement learning
19 Neither Precision Nor Architecture Alone: Controlled Tests of Failure Remedies for Physics-Informed Neural Networks 提出联合评估精度与架构以解决物理信息神经网络问题 SSM state space model

🔬 支柱八：物理动画 (Physics-based Animation) (3 篇)

#题目一句话要点标签🔗⭐
20 CardioFusion-AI: Robust ECG--PPG Fusion for Multimodal Physiological Monitoring Under Signal Degradation 提出CardioFusion-AI以解决ECG和PPG信号融合中的可靠性问题 PULSE multimodal
21 Learning Continuous Regional Temperature Fields with Lead-Time and Resolution Queries 提出连续时空温度预报模型以解决固定输出限制问题 spatiotemporal
22 Canalization Before Generalization: Grokking as a Dynamical Probe 提出Grokking机制以解决过参数化神经网络的泛化问题 PULSE

🔬 支柱一：机器人控制 (Robot Control) (1 篇)

#题目一句话要点标签🔗⭐
23 Beyond Pairwise Feedback: Listwise Vision-Language Supervision for Preference-Based Reward Learning 提出基于列表的视觉语言监督以提升偏好奖励学习效果 manipulation reinforcement learning RLHF

🔬 支柱七：动作重定向 (Motion Retargeting) (1 篇)

#题目一句话要点标签🔗⭐
24 Modeling spatio-temporal locality in multi-step forecasting of geo-referenced time series 提出SPALT模型以解决地理传感器多步预测问题 spatial relationship

🔬 支柱四：生成式动作 (Generative Motion) (1 篇)

#题目一句话要点标签🔗⭐
25 A Constitutive Markov Physics-Informed Neural Operator (MPNO) for Autoregressive Stability in Transient Dynamics 提出构成性马尔可夫物理信息神经算子以解决瞬态动力学中的自回归稳定性问题 penetration

⬅️ 返回 cs.LG 首页 · 🏠 返回主页