cs.LG（2025-05-27）

📊 共 42 篇论文 | 🔗 6 篇有代码

🎯 兴趣领域导航

支柱九：具身大模型 (Embodied Foundation Models) (23 🔗3) 支柱二：RL算法与架构 (RL & Architecture) (16 🔗2) 支柱一：机器人控制 (Robot Control) (1) 支柱四：生成式动作 (Generative Motion) (1 🔗1) 支柱八：物理动画 (Physics-based Animation) (1)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (23 篇)

#题目一句话要点标签🔗⭐
1 Let Me Think! A Long Chain-of-Thought Can Be Worth Exponentially Many Short Ones 长链思维胜过短链并行：揭示语言模型推理中序列计算的指数级优势 large language model chain-of-thought
2 DeCAF: Decentralized Consensus-And-Factorization for Low-Rank Adaptation of Foundation Models DeCAF：用于联邦学习中低秩适应的基础模型共识与分解算法 large language model foundation model
3 Multimodal Federated Learning: A Survey through the Lens of Different FL Paradigms 多模态联邦学习综述：从不同联邦学习范式的视角分析挑战与机遇 multimodal
4 LaX: Boosting Low-Rank Training of Foundation Models via Latent Crossing 提出LaX模块，通过潜在空间交叉提升低秩模型在预训练和微调中的性能 foundation model
5 Towards Robust Assessment of Pathological Voices via Combined Low-Level Descriptors and Foundation Model Representations 提出VOQANet+，结合底层声学特征与语音基础模型表征，提升病理嗓音评估的鲁棒性。 foundation model
6 Efficient Large Language Model Inference with Neural Block Linearization 提出神经块线性化（NBL）加速LLM推理，无需微调且精度损失小 large language model ✅
7 LLaMEA-BO: A Large Language Model Evolutionary Algorithm for Automatically Generating Bayesian Optimization Algorithms 提出LLaMEA-BO，利用大语言模型和进化算法自动生成贝叶斯优化算法。 large language model ✅
8 Generalizable Heuristic Generation Through Large Language Models with Meta-Optimization 提出MoH框架，利用LLM元优化启发式算法，提升组合优化问题泛化性。 large language model
9 PreGenie: An Agentic Framework for High-quality Visual Presentation Generation PreGenie：基于Agent框架的高质量可视化演示文稿生成 large language model multimodal
10 From Directions to Cones: Exploring Multidimensional Representations of Propositional Facts in LLMs 提出概念锥方法，探索LLM中命题事实的多维表示，提升真假判断干预效果。 large language model
11 PolarGrad: A Class of Matrix-Gradient Optimizers from a Unifying Preconditioning Perspective 提出PolarGrad，一种基于矩阵梯度极分解的矩阵梯度优化器，提升语言模型预训练效果。 large language model
12 MLE-STAR: Machine Learning Engineering Agent via Search and Targeted Refinement 提出MLE-STAR以解决机器学习工程代理的深度探索问题 large language model
13 Breaking AR's Sampling Bottleneck: Provable Acceleration via Diffusion Language Models 利用扩散语言模型加速AR采样：可证明的加速框架 large language model
14 Improving LLM-based Global Optimization with Search Space Partitioning HOLLM：通过搜索空间划分提升基于LLM的全局优化性能 large language model
15 Towards Interpretability Without Sacrifice: Faithful Dense Layer Decomposition with Mixture of Decoders 提出混合解码器(MxDs)，实现语言模型中稠密层的忠实且可解释分解。 large language model ✅
16 Pioneering 4-Bit FP Quantization for Diffusion Models: Mixup-Sign Quantization and Timestep-Aware Fine-Tuning 首创扩散模型4比特浮点量化：提出混合符号量化与时间步感知微调 large language model
17 Jailbreak-as-a-Service++: Unveiling Distributed AI-Driven Malicious Information Campaigns Powered by LLM Crowdsourcing PoisonSwarm：提出一种基于LLM众包的分布式恶意信息生成框架 large language model
18 Pause Tokens Strictly Increase the Expressivity of Constant-Depth Transformers 引入暂停符号显著提升恒定深度Transformer的表达能力 chain-of-thought
19 FireQ: Fast INT4-FP8 Kernel and RoPE-aware Quantization for LLM Inference Acceleration FireQ：面向LLM推理加速的快速INT4-FP8内核与RoPE感知量化 large language model
20 Convergence of Clipped-SGD for Convex $(L_0,L_1)$-Smooth Optimization with Heavy-Tailed Noise 针对重尾噪声下的凸(L0,L1)-光滑优化，提出Clipped-SGD收敛性保证 large language model
21 ChemHAS: Hierarchical Agent Stacking for Enhancing Chemistry Tools ChemHAS：通过层级代理堆叠增强化学工具性能，有效补偿预测误差。 large language model
22 'Hello, World!': Making GNNs Talk with LLMs 提出Graph Lingual Network (GLN)，利用LLM使GNN具备可解释性，并在图任务上取得良好零样本性能。 large language model
23 Can Past Experience Accelerate LLM Reasoning? 提出SpeedupLLM框架，加速LLM在重复任务上的推理速度并降低计算成本 large language model

🔬 支柱二：RL算法与架构 (RL & Architecture) (16 篇)

#题目一句话要点标签🔗⭐
24 TuneComp: Joint Fine-tuning and Compression for Large Foundation Models 提出TuneComp：联合微调与压缩大型基础模型，提升性能并减小模型体积。 distillation foundation model
25 A Cross Modal Knowledge Distillation & Data Augmentation Recipe for Improving Transcriptomics Representations through Morphological Features 提出跨模态知识蒸馏与数据增强方法，利用形态学特征提升转录组学表征 distillation foundation model multimodal
26 Foundation Model Hidden Representations for Heart Rate Estimation from Auscultation 利用预训练声学基础模型表征进行听诊心率估计，性能媲美甚至超越传统方法。 MAE foundation model
27 TabReason: A Reinforcement Learning-Enhanced Reasoning LLM for Explainable Tabular Data Prediction 提出TabReason，一种强化学习增强的推理LLM，用于可解释的表格数据预测。 reinforcement learning predictive model large language model
28 Deep Reinforcement Learning Agents are not even close to Human Intelligence HackAtari揭示深度强化学习智能体在简化任务中泛化能力不足 reinforcement learning deep reinforcement learning
29 Topology-Aware and Highly Generalizable Deep Reinforcement Learning for Efficient Retrieval in Multi-Deep Storage Systems 提出拓扑感知深度强化学习，用于多深度存储系统高效检索 reinforcement learning deep reinforcement learning
30 Hierarchical Reinforcement Learning with Uncertainty-Guided Diffusional Subgoals 提出基于不确定性引导扩散子目标的层级强化学习方法，提升样本效率和性能。 reinforcement learning diffusion policy
31 Simple yet Effective Graph Distillation via Clustering 提出ClustGDD，通过聚类实现高效图数据蒸馏，加速GNN训练。 representation learning distillation
32 Semi-supervised Clustering Through Representation Learning of Large-scale EHR Data 提出SCORE半监督聚类框架，通过表征学习处理大规模EHR数据，提升患者分型和预测能力。 predictive model representation learning
33 Accelerating RL for LLM Reasoning with Optimal Advantage Regression 提出A*-PO算法，通过最优优势回归加速LLM推理的强化学习训练。 reinforcement learning PPO large language model ✅
34 A Framework for Adversarial Analysis of Decision Support Systems Prior to Deployment 提出一种决策支持系统对抗分析框架，用于评估和防御深度强化学习智能体的安全风险。 reinforcement learning deep reinforcement learning DRL
35 Universal Value-Function Uncertainties 提出通用价值函数不确定性(UVU)方法，高效量化强化学习中的价值不确定性。 reinforcement learning offline RL distillation
36 HAD: Hybrid Architecture Distillation Outperforms Teacher in Genomic Sequence Modeling 提出混合架构蒸馏(HAD)，提升基因组序列建模中小模型性能，超越大模型教师。 distillation
37 A reinforcement learning agent for maintenance of deteriorating systems with increasingly imperfect repairs 提出基于强化学习的维护策略，解决退化系统日益不完善的维修问题。 reinforcement learning
38 Apprenticeship learning with prior beliefs using inverse optimization 利用逆优化的先验信念进行学徒学习，解决逆强化学习中的病态问题。 reinforcement learning inverse reinforcement learning
39 Sparsified State-Space Models are Efficient Highway Networks 提出Simba方法以提高状态空间模型的效率 Mamba SSM ✅

🔬 支柱一：机器人控制 (Robot Control) (1 篇)

#题目一句话要点标签🔗⭐
40 Multi-level Certified Defense Against Poisoning Attacks in Offline Reinforcement Learning 提出多层认证防御机制，提升离线强化学习抵抗投毒攻击的鲁棒性 manipulation reinforcement learning offline RL

🔬 支柱四：生成式动作 (Generative Motion) (1 篇)

#题目一句话要点标签🔗⭐
41 Conditional Diffusion Models with Classifier-Free Gibbs-like Guidance 提出无分类器引导的吉布斯采样以解决扩散模型样本多样性问题 classifier-free guidance ✅

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

#题目一句话要点标签🔗⭐
42 Towards Operational Automated Greenhouse Gas Plume Detection 利用卷积神经网络实现温室气体羽流的自动化检测与运营部署 spatiotemporal

⬅️ 返回 cs.LG 首页 · 🏠 返回主页