cs.LG（2025-01-27）

📊 共 34 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) (17 🔗3) 支柱二：RL算法与架构 (RL & Architecture) (16 🔗3) 支柱三：空间感知与语义 (Perception & Semantics) (1)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (17 篇)

#题目一句话要点标签🔗⭐
1 LLM-attacker: Enhancing Closed-loop Adversarial Scenario Generation for Autonomous Driving with Large Language Models LLM-attacker：利用大语言模型增强自动驾驶闭环对抗场景生成 large language model
2 Rethinking the Bias of Foundation Model under Long-tailed Distribution 针对长尾分布下预训练模型偏差，提出基于因果学习的解耦方法。 foundation model ✅
3 SIM: Surface-based fMRI Analysis for Inter-Subject Multimodal Decoding from Movie-Watching Experiments 提出基于表面视觉Transformer的SIM模型，实现跨个体多模态脑解码，提升电影观看实验中的泛化能力。 multimodal ✅
4 Zero-Shot Decision Tree Construction via Large Language Models 提出一种基于大语言模型的零样本决策树构建方法，无需训练数据。 large language model
5 Phase Transitions in Large Language Models and the $O(N)$ Model 将Transformer架构重构为O(N)模型，揭示大语言模型中的相变现象 large language model
6 LemmaHead: RAG Assisted Proof Generation Using Large Language Models LemmaHead：利用RAG辅助大型语言模型生成数学证明 large language model
7 Distributional Information Embedding: A Framework for Multi-bit Watermarking 提出分布信息嵌入框架，用于大语言模型多比特水印 large language model
8 Smoothed Embeddings for Robust Language Models 提出RESTA防御方法，通过平滑嵌入向量增强语言模型对抗恶意攻击的鲁棒性。 large language model
9 CoCoNUT: Structural Code Understanding does not fall out of a tree CoCoNUT：揭示大型语言模型在代码结构理解方面的局限性 large language model
10 Detecting Zero-Day Attacks in Digital Substations via In-Context Learning 提出基于上下文学习的零日攻击检测方法，用于保障数字变电站安全 large language model
11 Language-Based Bayesian Optimization Research Assistant (BORA) 提出基于语言模型的贝叶斯优化助手BORA，融合领域知识提升复杂实验优化效率。 large language model
12 SWIFT: Mapping Sub-series with Wavelet Decomposition Improves Time Series Forecasting 提出SWIFT模型，利用小波分解提升轻量级时间序列预测在边缘计算中的性能。 large language model ✅
13 TimeHF: Billion-Scale Time Series Models Guided by Human Feedback TimeHF：基于人类反馈的十亿级时间序列模型，提升供应链预测精度。 large language model
14 SkillScope: A Tool to Predict Fine-Grained Skills Needed to Solve Issues on GitHub SkillScope：预测GitHub问题所需细粒度技能，辅助开源项目贡献者。 large language model
15 Investigating the Sensitivity of Pre-trained Audio Embeddings to Common Effects 研究预训练音频嵌入对常见音频效果的敏感性，揭示其鲁棒性局限 foundation model
16 Adaptive Width Neural Networks 提出自适应宽度神经网络，通过反向传播联合优化网络宽度和参数。 foundation model
17 GraphICL: Unlocking Graph Learning Potential in LLMs through Structured Prompt Design GraphICL：通过结构化提示设计释放LLM在图学习中的潜力 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🔗⭐
18 Objects matter: object-centric world models improve reinforcement learning in visually complex environments 提出OC-STORM，利用对象中心世界模型提升视觉复杂环境中强化学习的样本效率。 reinforcement learning deep reinforcement learning world model
19 Inverse Reinforcement Learning via Convex Optimization 提出基于凸优化的逆强化学习方法，提升鲁棒性和可复现性。 reinforcement learning inverse reinforcement learning
20 Mixture-of-Mamba: Enhancing Multi-Modal State-Space Models with Modality-Aware Sparsity 提出Mixture-of-Mamba以解决多模态状态空间模型的稀疏性问题 Mamba SSM state space model ✅
21 Upside Down Reinforcement Learning with Policy Generators 提出基于策略生成器的倒置强化学习（UDRLPG）框架，提升强化学习样本效率。 reinforcement learning multimodal ✅
22 Application of Structured State Space Models to High energy physics with locality-sensitive hashing 提出基于局部敏感哈希的结构化状态空间模型，用于解决高能物理领域长序列处理难题。 Mamba SSM state space model
23 sDREAMER: Self-distilled Mixture-of-Modality-Experts Transformer for Automatic Sleep Staging 提出sDREAMER模型，利用自蒸馏混合模态专家Transformer进行自动睡眠分期 dreamer distillation
24 Towards General-Purpose Model-Free Reinforcement Learning 提出MR.Q算法，通过模型表示线性化值函数，实现通用无模型强化学习。 reinforcement learning model-based RL
25 Training Dynamics of In-Context Learning in Linear Attention 研究线性注意力中上下文学习的训练动态，揭示参数化方式对学习过程的影响 linear attention
26 The Effect of Optimal Self-Distillation in Noisy Gaussian Mixture Model 研究噪声高斯混合模型中自蒸馏的有效性，揭示其去噪机制并提出优化策略。 distillation
27 ReFill: Reinforcement Learning for Fill-In Minimization ReFill：提出基于强化学习的填充最小化方法，提升稀疏线性系统求解效率 reinforcement learning
28 Multi-Objective Reinforcement Learning for Power Grid Topology Control 提出基于多目标强化学习的电网拓扑控制方法，优化线路负载和拓扑结构。 reinforcement learning
29 Efficient Logit-based Knowledge Distillation of Deep Spiking Neural Networks for Full-Range Timestep Deployment 提出高效的基于Logit的知识蒸馏方法以解决深度脉冲神经网络的时间步部署问题 distillation ✅
30 The Sample Complexity of Online Reinforcement Learning: A Multi-model Perspective 提出在线强化学习样本复杂度分析方法以应对非线性动态系统 reinforcement learning
31 Benchmarking Quantum Reinforcement Learning 提出一种量子强化学习的基准测试方法，用于评估和验证量子算法的性能。 reinforcement learning
32 Foundation for unbiased cross-validation of spatio-temporal models for species distribution modeling 提出基于空间自相关的交叉验证方法，提升物种分布模型时空泛化能力。 SAC MAE
33 Challenging Assumptions in Learning Generic Text Style Embeddings 提出基于对比学习的通用文本风格嵌入方法，并反思现有假设 representation learning contrastive learning

🔬 支柱三：空间感知与语义 (Perception & Semantics) (1 篇)

#题目一句话要点标签🔗⭐
34 INRet: A General Framework for Accurate Retrieval of INRs for Shapes INRet：用于精确检索形状INR的通用框架，支持多种INR架构和隐式函数。 scene reconstruction

⬅️ 返回 cs.LG 首页 · 🏠 返回主页