cs.LG（2026-10-08）

📊 共 57 篇论文 | 🔗 8 篇有代码

🎯 兴趣领域导航

支柱二：RL算法与架构 (RL & Architecture) (29 🔗3) 支柱九：具身大模型 (Embodied Foundation Models) (19 🔗3) 支柱一：机器人控制 (Robot Control) (4 🔗2) 支柱八：物理动画 (Physics-based Animation) (2) 支柱五：交互与反应 (Interaction & Reaction) (2) 支柱四：生成式动作 (Generative Motion) (1)

🔬 支柱二：RL算法与架构 (RL & Architecture) (29 篇)

#题目一句话要点标签🔗⭐
1 CausalDreamer: Learning Predictive World Models with Latent Disentanglement 提出CausalDreamer以解决生成世界模型的可控性与奖励相关性问题 world model world models dreamer
2 SpatialOPSD: Self-Distilling Spatial Intelligence from Verified Coding Agent Traces 提出SpatialOPSD以解决空间推理的工具依赖问题 distillation privileged information large language model
3 ActiveMedAgent: Cost-Aware Trajectory Learning for Multimodal Medical Diagnosis 提出ActiveMedAgent以解决多模态医疗诊断中的成本意识问题 policy learning multimodal
4 Poster: A Preliminary Study of LLM Distillation Inference 提出蒸馏推理方法以检测模型蒸馏攻击 distillation large language model
5 GRPODropout: Less is More for Online Reinforcement Learning Rollouts 提出GRPODropout以解决强化学习中的策略熵崩溃问题 reinforcement learning large language model ✅
6 Can Jev be Your Q or Policy in Reinforcement Learning? 提出Jev作为强化学习中的决策模型以提升样本效率 reinforcement learning foundation model
7 Policy Alignment: New Signals for Membership Auditing in On-Policy Distillation 提出政策对齐会员审计框架以解决现有方法不足 teacher-student distillation
8 Why On-Policy Distillation Sometimes Fails: Vanishing Learning Signals 分析在策略蒸馏中信号消失问题及其解决方案 reinforcement learning distillation ✅
9 Do LLMs Learn from Rewards in Context? : Rethinking the role of reward in In-Context Reinforcement Learning 研究LLM在上下文中如何学习奖励，重新思考奖励在上下文强化学习中的角色 reinforcement learning reward shaping
10 SCORE: Spectral Correlation Estimation for Multivariate Gaussians 提出SCORE框架以高效估计多元高斯的协方差矩阵 predictive model depth estimation monocular depth
11 Bi-FORK: Generative Modeling of High-Dimensional Bifurcating Systems 提出Bi-FORK以解决高维分岔系统的生成建模问题 flow matching multimodal
12 A Unified Bellman Operator for Safety-Critical Reinforcement Learning 提出统一贝尔曼算子以解决安全关键强化学习问题 reinforcement learning
13 Is Real-World Training Data Necessary for Generalist Graph Anomaly Detection? 提出AG-FORGE与TS-GGAD以解决图异常检测数据不足问题 curriculum learning foundation model
14 Softmax Attention on Gaussian Mixtures: Linear When It Can, Selective When It Must 提出高斯混合模型下的Softmax注意力以解决非线性任务 linear attention multimodal
15 $C_4$-Equivariant Flow Matching on Anisotropic Power-Diagram Graphs for Microstructure Generation 提出基于$C_4$对称流匹配的微观结构生成方法 flow matching
16 From a Prompt to Repertoires: Evolving Functional REpertoires Enable LLM Continual Learning 提出Evolving Functional REpertoires以解决大语言模型的持续学习问题 reinforcement learning large language model
17 Residual spectral instabilities in representation learning 提出基于波动理论的变分自编码器后验崩溃分析方法 representation learning
18 Bridging KV-Cache Quantization and Linear Attention: From Theory to Pretrained Weight Migration 提出RAM-Net以桥接KV缓存量化与线性注意力问题 linear attention
19 Few-Step Generation via Data-Space Iteration 提出数据空间迭代以解决生成模型的采样效率问题 flow matching distillation
20 Credal Machine Learning for Risk-Averse Decision Making 提出基于信念集的机器学习方法以解决风险规避决策问题 reinforcement learning predictive model
21 NV-Tesseract-AD: Curriculum-Enhanced Conditional Diffusion for Multivariate Time-Series Anomaly Detection 提出NV-Tesseract-AD以解决多变量时间序列异常检测问题 representation learning curriculum learning zero-shot transfer
22 Cross-species representation learning aligns mouse and human neural dynamics and tracks clinical drug efficacy 提出跨物种神经表征学习以解决药物临床效果预测问题 representation learning contrastive learning
23 PIVOT: Perplexity-Informed KD-to-RL Transition Scheduling for Vertical-Domain Few-Shot Distillation 提出PIVOT以解决小语言模型在垂直领域少样本分类中的知识获取问题 reinforcement learning distillation
24 DaCe-DT: Data-Centric Offline Multi-Task Reinforcement Learning via Adaptive Prompts and Trajectory Correction for Heterogeneous Tasks 提出DaCe-DT以解决离线多任务强化学习中的数据瓶颈问题 reinforcement learning policy learning
25 Primal-Dual Flow Matching for Sample-Wise Constrained Generation 提出原始-对偶流匹配方法以解决样本约束生成问题 flow matching ✅
26 Toward Maximizing Algebraic Connectivity via Density-Dependent Backbone Construction and Reinforcement Learning Completion 提出基于密度依赖的骨干构建和强化学习的代数连通性最大化方法 reinforcement learning
27 Emergent Inverse-Depth Scaling From Nonlinearity In Attention 提出非线性注意力机制以改善模型深度缩放问题 linear attention large language model
28 KITE: A Knowledge-Injected Temporal Encoder for Time-Series Event Forecasting 提出KITE以解决多变量时间序列事件预测中的知识注入问题 world model world models
29 SACQ: Structured Decoding with Memory-Conditioned Refinement for Long-Horizon Forecasting 提出SACQ以解决长时间序列预测中的历史与未来对齐问题 Mamba MAE

🔬 支柱九：具身大模型 (Embodied Foundation Models) (19 篇)

#题目一句话要点标签🔗⭐
30 Automated Assembly Instruction Generation from CAD Models Using Grounded Large Language Models: A Human-in-the-Loop Framework 提出基于CAD模型的自动化装配指令生成框架以提升装配文档质量 large language model language conditioned
31 Multimodal Graph Retrieval-Augmented Sequential Recommendation via Collaborative Filtering Paths 提出MGRASRec以解决多模态推荐中的协同信号缺失问题 large language model multimodal
32 CoPoE: Multimodal Fusion via Decomposable Disease-Coordinate Product-of-Experts for Missing-Modality Alzheimer's Diagnosis 提出CoPoE以解决阿尔茨海默病多模态诊断中的缺失模态问题 multimodal ✅
33 Deflating the Hessian: Rank-4 W4A4 Quantization for Multimodal Diffusion Transformers 提出统一框架以解决多模态扩散变换器的量化损失问题 multimodal
34 Test-Time Compute for Tabular Foundation Models: Mechanisms, Gains, and Limits 提出测试时计算机制以提升表格基础模型的预测性能 foundation model ✅
35 Timer-M1: A Multivariate Time Series Foundation Model via Learning Primitives 提出Timer-M1以解决多变量时间序列零-shot预测问题 foundation model
36 Addressing Overcommitment in the Reasoning of Gendered Economic Memes under Multimodal Ambiguity 提出CGER-Net以解决多模态经济表述中的性别偏见问题 multimodal
37 SparseDecoding: Decoding-Aware Pruning for Accurate and Efficient LLM Inference 提出SparseDecoding以解决大语言模型推理中的延迟问题 large language model
38 AdaCast: Conditional Parameter Generation for Adaptive Time Series Forecasting 提出AdaCast以解决时间序列预测中的适应性问题 foundation model
39 $σ$Transfer: Uncertainty Transfer from Small to Large Networks under $μ\mathrm{P}$ 提出$σ$Transfer以实现小模型到大模型的不确定性转移 zero-shot transfer
40 Deflating the Hessian: Rank-4 W4A4 Quantization for Multimodal Diffusion Transformers 提出HSVDQuant以解决4位量化损失问题 multimodal
41 A Distributional Framework for Behavioral Characterization and Risk Management of Large Language Models 提出分布式框架以管理大型语言模型的行为与风险 large language model
42 Byzantine-Resilient Federated Large Language Model via Hypergraph Message Passing 提出超图消息传递框架以解决联邦大语言模型的拜占庭攻击问题 large language model
43 Machine Learning and Time-Series Foundation Models for Day-Ahead Electricity Forecasting in a University Building 提出基于机器学习的模型以优化大学建筑的电力预测 foundation model
44 Market Impact Estimation via Causal Foundation Models 提出因果市场模拟器以解决市场影响估计问题 foundation model
45 Meta-Reflective Evolution Effectively Improves Production AI Systems 提出CoCoEvolve以优化生产AI系统的演化方法 large language model
46 REDaCT: Collapse-Free Unlearning in Large Reasoning Models 提出REDaCT以解决大规模推理模型中的遗忘问题 chain-of-thought ✅
47 Sample-Adaptive Zeroth-Order Post-Training 提出样本自适应零阶后训练方法以提升收敛速度 large language model
48 RideBench: A Large-Scale Exogenous-Aware Benchmark for Ride-Hailing Time Series Forecasting 提出RideBench以解决出行打车时间序列预测中的外部因素影响问题 foundation model

🔬 支柱一：机器人控制 (Robot Control) (4 篇)

#题目一句话要点标签🔗⭐
49 When to Intervene? State-Aware Sparse Manipulation in Federated Reinforcement Learning 提出V-BSA以解决FRL中的干预时机问题 manipulation reinforcement learning policy learning ✅
50 A Geometric Approach to Soft Actor-Critic with Zonotopes for Locomotion Learning 提出GeZo-SAC以解决演员-评论家方法中的过估计偏差问题 locomotion SAC
51 SplitJEPA: Learning Invariant and Variant Latent Worlds without Reconstruction 提出SplitJEPA以解决动态世界中不变与变异潜在结构学习问题 manipulation JEPA Joint-Embedding Predictive Architecture
52 When to Intervene? State-Aware Sparse Manipulation in Federated Reinforcement Learning 提出V-BSA以解决FRL中的干预时机问题 manipulation reinforcement learning policy learning ✅

🔬 支柱八：物理动画 (Physics-based Animation) (2 篇)

#题目一句话要点标签🔗⭐
53 LAIR-Net: Leaky Alignment-Impulse Residual Networks for Tabular Regression 提出LAIR-Net以解决表格回归中的非线性目标学习问题 PULSE
54 A Graph Neural Network for Global Daily Fire Radiative Power Prediction at Medium-Range Lead Times 提出图神经网络以解决全球火灾辐射功率预测问题 spatiotemporal

🔬 支柱五：交互与反应 (Interaction & Reaction) (2 篇)

#题目一句话要点标签🔗⭐
55 Agentic-TTT: Training test-time policy for test-time training 提出Agentic-TTT以解决测试时训练中的决策问题 IMoS
56 PVBRF-ID: Private, Verifiable and Byzantine-Robust Federated Intrusion Detection for Oilfield Enterprise Networks 提出PVBRF-ID以解决油田企业网络的入侵检测隐私与鲁棒性问题 OMOMO

🔬 支柱四：生成式动作 (Generative Motion) (1 篇)

#题目一句话要点标签🔗⭐
57 Transformer-Based Masked Diffusion Models Provably Learn Distributions with Latent Structure 提出基于Transformer的掩蔽扩散模型以解决分布学习问题 MDM

⬅️ 返回 cs.LG 首页 · 🏠 返回主页