cs.LG（2025-03-05）

📊 共 29 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) (14 🔗2) 支柱二：RL算法与架构 (RL & Architecture) (9 🔗1) 支柱八：物理动画 (Physics-based Animation) (3) 支柱七：动作重定向 (Motion Retargeting) (1) 支柱三：空间感知与语义 (Perception & Semantics) (1) 支柱四：生成式动作 (Generative Motion) (1)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (14 篇)

#题目一句话要点标签🔗⭐
1 Multimodal Stock Price Prediction: A Case Study of the Russian Securities Market 提出一种融合新闻文本和时间序列的多模态方法，用于提升俄罗斯股市价格预测精度。 large language model multimodal
2 LLM as GNN: Graph Vocabulary Learning for Text-Attributed Graph Foundation Models 提出PromptGFM，通过图词汇学习实现文本属性图的图基础模型。 large language model foundation model ✅
3 Exploring the Potential of Large Language Models as Predictors in Dynamic Text-Attributed Graphs 提出GraphAgent-Dynamic框架，利用协作LLM解决动态文本属性图预测难题。 large language model foundation model
4 TEDDY: A Family Of Foundation Models For Understanding Single Cell Biology TEDDY：用于理解单细胞生物学的系列Transformer基础模型 foundation model
5 PAIR: A Novel Large Language Model-Guided Selection Strategy for Evolutionary Algorithms PAIR：基于大语言模型引导的进化算法选择策略，提升TSP问题求解性能 large language model
6 Bridging Molecular Graphs and Large Language Models 提出Graph2Token，将分子图与大语言模型对齐，实现分子性质预测。 large language model
7 An Optimization Algorithm for Multimodal Data Alignment 提出AlignXpert算法，优化多模态数据对齐，提升跨模态推理能力 multimodal
8 LLM-Powered Prediction of Hyperglycemia and Discovery of Behavioral Treatment Pathways from Wearables and Diet 提出GlucoLens以预测餐后高血糖并发现行为治疗路径 large language model multimodal ✅
9 LEWIS (LayEr WIse Sparsity) -- A Training Free Guided Model Merging Approach LEWIS：一种免训练的层级稀疏引导模型合并方法 large language model instruction following
10 A Little Depth Goes a Long Way: The Expressive Power of Log-Depth Transformers 提出对数深度Transformer，解决传统Transformer在长序列推理上的表达能力不足问题 chain-of-thought
11 The MASK Benchmark: Disentangling Honesty From Accuracy in AI Systems 提出MASK基准，用于区分AI系统中的诚实性与准确性 large language model
12 Memory Injection Attacks on LLM Agents via Query-Only Interaction 提出MINJA：一种针对LLM Agent的查询注入式记忆攻击方法 large language model
13 Mixture of Experts Made Intrinsically Interpretable 提出MoE-X，一种本质上可解释的混合专家语言模型，提升模型可解释性。 large language model
14 Robust Learning of Diverse Code Edits 提出SeleKT算法和NextCoder模型，提升代码语言模型在多样化代码编辑任务中的鲁棒性。 instruction following

🔬 支柱二：RL算法与架构 (RL & Architecture) (9 篇)

#题目一句话要点标签🔗⭐
15 Feature Matching Intervention: Leveraging Observational Data for Causal Representation Learning 提出特征匹配干预(FMI)方法，从观测数据中学习因果表示 representation learning feature matching
16 Rebalanced Multimodal Learning with Data-aware Unimodal Sampling 提出数据感知的单模态采样方法，解决多模态学习中的模态不平衡问题。 reinforcement learning multimodal
17 Seldonian Reinforcement Learning for Ad Hoc Teamwork 提出基于Seldonian优化的离线强化学习方法，解决Ad Hoc团队合作中的安全可靠性问题 reinforcement learning offline RL
18 State-offset Tuning: State-based Parameter-Efficient Fine-Tuning for State Space Models 提出State-offset Tuning，一种面向状态空间模型的高效参数微调方法 SSM state space model ✅
19 Towards Understanding Distilled Reasoning Models: A Representational Approach 通过表征分析理解蒸馏推理模型：揭示模型蒸馏对LLM推理能力的影响 distillation large language model
20 Probabilistic Insights for Efficient Exploration Strategies in Reinforcement Learning 针对稀疏奖励强化学习，提出基于概率洞察的高效探索策略 reinforcement learning
21 Chunking the Critic: A Transformer-based Soft Actor-Critic with N-Step Returns 提出基于Transformer的序列条件评论家，结合N步回报，提升SAC在长程任务中的性能。 reinforcement learning SAC
22 DO-IQS: Dynamics-Aware Offline Inverse Q-Learning for Optimal Stopping with Unknown Gain Functions 提出DO-IQS，解决未知增益函数下最优停止的逆向Q学习问题 reinforcement learning inverse reinforcement learning
23 Less is more? Rewards in RL for Cyber Defence 提出基于稀疏奖励的强化学习网络防御方法，提升复杂网络环境下的防御效果。 reinforcement learning deep reinforcement learning

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

#题目一句话要点标签🔗⭐
24 A Survey of Foundation Models for Environmental Science 综述：环境科学领域的基础模型应用与发展 spatiotemporal foundation model
25 Limits of nonlinear and dispersive fiber propagation for an optical fiber-based extreme learning machine 基于光纤传播的极限学习机：研究非线性与色散效应对性能的限制 PULSE
26 WVEmbs with its Masking: A Method For Radar Signal Sorting 提出基于掩码的宽值嵌入方法WVEmbs，用于复杂雷达信号分选。 PULSE

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

#题目一句话要点标签🔗⭐
27 Dimensionality reduction for homological stability and global structure preservation 提出DiRe工具包以解决传统降维方法的局限性 structure preservation

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

#题目一句话要点标签🔗⭐
28 Is Pre-training Applicable to the Decoder for Dense Prediction? 提出$ imes$Net，实现预训练解码器在密集预测任务中的应用，显著提升性能。 depth estimation monocular depth

🔬 支柱四：生成式动作 (Generative Motion) (1 篇)

#题目一句话要点标签🔗⭐
29 Handling Uncertainty in Health Data using Generative Algorithms RIGA：利用生成对抗网络处理医疗数据中的不确定性和类别不平衡问题 VQ-VAE

⬅️ 返回 cs.LG 首页 · 🏠 返回主页