cs.LG（2026-10-22）

📊 共 59 篇论文 | 🔗 3 篇有代码

🎯 兴趣领域导航

支柱九：具身大模型 (Embodied Foundation Models) (29) 支柱二：RL算法与架构 (RL & Architecture) (28 🔗3) 支柱四：生成式动作 (Generative Motion) (1) 支柱八：物理动画 (Physics-based Animation) (1)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (29 篇)

#题目一句话要点标签🔗⭐
1 Causal Chain of Thought: Mitigating Attention Leakage via Context Isolation 提出CausalCoT以解决注意力泄漏问题 chain-of-thought
2 CRANE: Post-Hoc Counterfactual Retrieval for Multimodal Classifiers 提出CRANE以解决多模态分类器的后验反事实检索问题 multimodal
3 Drift Says When, Budget Says How Much: Rate-Controlled KV Re-selection for Long-Form Multimodal Generation 提出基于漂移控制的KV重选方法以提升长文本多模态生成性能 multimodal
4 Track finding and particle identification performance with particle physics foundation models 提出基础模型以提升粒子物理数据分析性能 foundation model
5 Random Initialization as a Baseline for Time Series Foundation Models 提出随机初始化作为时间序列基础模型的基线评估方法 foundation model
6 Rethinking Sequence Compression with Test-time Regression 提出TTR-Compressor以解决序列压缩效率问题 large language model multimodal
7 Reliable AI-Assisted Decisions via Selective Auditing 提出选择性审计方法以提高AI决策的可靠性 large language model
8 Cost-Aware Abstention for LLM-Based IT Ticket Classification 提出成本感知的放弃策略以优化IT工单分类 large language model
9 Predictable performance is suboptimal, and what it means for model routing 提出性能预测模型以优化语言模型路由 large language model
10 SQuAT: Spherical Lattice Vector Quantization-Aware Training 提出SQuAT以提升大语言模型的量化训练效果 large language model
11 Metric Conformal Prediction via the Conditional Expected Local Radius 提出条件期望局部半径以解决非欧几里得响应的预测问题 multimodal
12 Harmful Responses Are Jailbreak Templates: Transferring Autoregressive Jailbreaks to Diffusion Language Models 提出RMT-Jailbreak以解决扩散语言模型的越狱攻击问题 large language model
13 DISC: Domain-Integrated Similarity Cost for Time-Series Dataset 提出DISC框架以优化时间序列数据集相似性评估 foundation model
14 Transformers with CoT Discover Heavy-Ball Acceleration Through Training 提出重球法加速的变换器训练方法 chain-of-thought
15 ReDence: A Hybrid Agentic and Decision-Model Framework for Rapid, Evidence-Grounded Detection of Modern Social Engineering Scams 提出ReDence框架以解决现代社交工程诈骗检测问题 multimodal
16 Amplifying Latency and Cost of LLM Inference via Rowhammer Attack 提出LLVMhammer攻击以降低LLM推理性能并增加成本 large language model
17 Gandalf: Flash Shall Not Die! Increasing Flash Storage Lifetime for KV Cache Offloading in LLM Serving 提出Gandalf以提高SSD在LLM服务中的使用寿命 large language model
18 Federated Graph Retrieval-Augmented Generation 提出联邦图检索增强生成以解决知识分布问题 large language model
19 Interpreting Model Reasoning through Behavioral Counterfactuals 提出行为反事实方法以解析模型推理过程 chain-of-thought
20 Autoregressive Multimodal Anomaly Detection with Large Language Models for Microservice Systems 提出LLMMAD框架以解决微服务系统中的多模态异常检测问题 large language model multimodal
21 Efficient Omni-modal Large Language Model Inference via Per-Query Modality Selection 提出基于查询的模态选择以优化Omni-LLM推理效率 large language model multimodal
22 Forecasting Associations, Not Interventions: A Preregistered Causal Audit of Covariate-Aware Time Series Foundation Models 提出因变量预测模型以解决干预效应识别问题 foundation model
23 ConsMSA: Semantic Distribution Consistency Learning for Multimodal Sentiment Analysis 提出ConsMSA以解决多模态情感分析中的语义一致性问题 multimodal
24 Track finding and particle identification performance with particle physics foundation models 提出基础模型以提升粒子物理数据分析性能 foundation model
25 TRACE: Data-Free Text Reconstruction Attacks against Approximate Unlearning in LLMs 提出TRACE框架以解决LLMs近似遗忘中的隐私泄露问题 large language model
26 Every Weight Is a Path of Shared Discrete Decisions 提出TWIG以优化神经网络权重表示与决策共享 large language model
27 Speculative Diffusion is a Tale of Two Tradeoffs 提出一种新的投机扩散方法以优化生成模型的速度与质量 large language model
28 When and How Does Muon Precondition? The Role of Structured Gradient Noise 研究Muon优化器在大语言模型训练中的预条件作用 large language model
29 Harmful Responses Are Jailbreak Templates: Transferring Autoregressive Jailbreaks to Diffusion Language Models 提出RMT-Jailbreak以解决扩散语言模型的越狱攻击问题 large language model

🔬 支柱二：RL算法与架构 (RL & Architecture) (28 篇)

#题目一句话要点标签🔗⭐
30 Inside-JEPA: From Prediction to Relational Grounding in Latent World Models 提出Inside-JEPA以解决潜在世界模型关系基础问题 world model world models JEPA
31 When News Gets Old: Portfolio Allocation with Reinforcement Learning and Large Language Models 提出基于强化学习和大语言模型的投资组合分配方法 reinforcement learning large language model
32 GIFT: Gradient Informed Filtering for Distillation Augmented Reasoning RL 提出GIFT以解决蒸馏与强化学习目标不兼容问题 reinforcement learning distillation large language model ✅
33 Fast Iterative Combinatorial Auctions via Deep Reinforcement Learning 提出FastCombo以解决迭代组合拍卖中的价格发现问题 reinforcement learning deep reinforcement learning
34 Future of Time Series Anomaly Detection: From Historical Review to Unified Outlook 提出统一框架以解决时间序列异常检测的多样性问题 state space model large language model zero-shot transfer
35 OP-Print: Tracing Membership Imprints in On-Policy Self-Distillation 提出OP-Print以解决自蒸馏中的隐私风险问题 teacher-student distillation
36 Ground-Truth-First Benchmarking of Reinforcement Learning via Converse Optimality: From Smooth, Fully Observed Systems to Non-Smooth or Partially Observed Tasks 提出基于反向最优性的基准测试框架以解决强化学习评估问题 reinforcement learning reward design
37 METTA: Memory-Enhanced Test-Time Adaptation for Meta-Reinforcement Learning 提出METTA以解决元强化学习中的快速适应问题 reinforcement learning
38 When to Follow the Teacher: Advantage-Based On-Policy Distillation 提出优势基础的在线策略蒸馏方法以提升学生学习效果 distillation
39 PriMa-Causa: Causal Prescriptive Maintenance 提出PriMa-Causa以解决制造业中的因果性预防性维护问题 predictive model foundation model
40 Building Presentation Intelligence for High-Quality Slide Generation 提出统一训练框架以实现高质量PPT自动生成 reinforcement learning large language model
41 Tracing Subliminal Learning Through Pretraining 研究潜在偏好传递机制以优化模型蒸馏训练 distillation large language model
42 One Teacher, Many Views: Reducing Reference Sensitivity in On-Policy Self-Distillation 提出蒙特卡洛特权蒸馏以解决参考敏感性问题 distillation
43 MOER: Mix-Policy Optimization under Expert Prefix Guidance with Selective Rollback 提出MOER框架以解决强化学习中的专家指导与自我探索问题 reinforcement learning chain-of-thought ✅
44 Reverse-KL Policy Distillation: Statistical Guarantees and Limits 提出反KL蒸馏方法以解决统计保证与限制问题 distillation
45 Primal-Dual Representation Learning for Low-Rank Constrained MDPs 提出REP-PD-TEG以解决低秩约束MDP中的表示学习问题 representation learning
46 Learning Domain Trajectories with Flow Matching for Gradual Domain Adaptation 提出FlowGDA框架以解决渐进式领域适应中的分布偏移问题 flow matching
47 An Agent Architecture with Tools and Memory for Selective Correction of Wrist PPG Heart Rate Estimates 提出一种新型代理架构以解决腕部PPG心率估计的选择性修正问题 MAE PULSE
48 Questioning Instrumental Convergence in RLVR: Checkpoint Evidence from OLMo-3 研究RLVR中的工具性收敛现象及其对模型行为的影响 reinforcement learning DPO
49 Aligning Large Language Models with Human Feedback: Mathematical Foundations and Algorithm Design 提出数学基础与算法设计以对齐大型语言模型与人类反馈 reinforcement learning inverse reinforcement learning RLHF
50 Learning to Be Monitored: Causal Bounds and Intervention-Aware Post-Training Recipes for Chain-of-Thought Monitorability 提出干预感知的后训练方法以提升链式思维监控能力 reinforcement learning distillation large language model ✅
51 Does LeWorldModel Preserve Local Action Geometry? 评估LeWorldModel在局部动作几何保留上的表现 world model worldmodel world models
52 Beyond Local Accuracy: Benchmarking the Rollout Reliability of PDE Foundation Models 提出PDE基础模型的长期可靠性评估方法 MAE foundation model
53 Effectiveness of Curriculum Learning Depends on Reward Sparsity and Competing Optima 提出课程学习方法以解决稀疏奖励与竞争最优问题 reinforcement learning policy learning curriculum learning
54 Compatible Likelihoods for Flow Matching on Manifolds 提出兼容性似然以优化流匹配问题 flow matching
55 Rethinking Multi-View Fusion for Urban Representation Learning 提出PRISM框架以解决城市区域表示学习中的多视角融合问题 representation learning
56 REDAR: Reinforcement-Driven Adaptive Regularisation for Class Incremental Learning 提出REDAR以解决类增量学习中的正则化选择问题 reinforcement learning distillation
57 Cognitively Constrained Partners for Zero-Shot Coordination in Real-Time Partitioned Control 提出认知约束伙伴以解决实时协调问题 reinforcement learning reward shaping

🔬 支柱四：生成式动作 (Generative Motion) (1 篇)

#题目一句话要点标签🔗⭐
58 Random Remasking Scales Masked Diffusion Inference 提出随机重掩码方法以提升掩码扩散推理效果 MDM

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

#题目一句话要点标签🔗⭐
59 End-to-End Differentiable Diffusion Conditioning for Physics-Informed Optimization 提出可微分扩散条件优化方法以解决物理系统优化问题 PULSE

⬅️ 返回 cs.LG 首页 · 🏠 返回主页