cs.LG（2025-02-04）

📊 共 15 篇论文

🎯 兴趣领域导航

支柱九：具身大模型 (Embodied Foundation Models) (8) 支柱二：RL算法与架构 (RL & Architecture) (4) 支柱一：机器人控制 (Robot Control) (2) 支柱三：空间感知与语义 (Perception & Semantics) (1)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (8 篇)

#题目一句话要点标签🔗⭐
1 Federated Low-Rank Tensor Estimation for Multimodal Image Reconstruction 提出基于联邦学习的低秩张量估计方法，用于多模态图像重建。 multimodal
2 Shuttle Between the Instructions and the Parameters of Large Language Models 提出SHIP框架，学习大语言模型指令与参数间的双向映射关系 large language model
3 Vision-Language Model Dialog Games for Self-Improvement 提出VLM对话游戏自提升框架，解决视觉-语言模型训练数据瓶颈问题 multimodal
4 Peri-LN: Revisiting Normalization Layer in the Transformer Architecture 提出Peri-LN，一种新型Transformer归一化策略，提升大规模模型训练稳定性和收敛速度。 large language model
5 Twilight: Adaptive Attention Sparsity with Hierarchical Top-$p$ Pruning Twilight：利用分层Top-$p$剪枝实现自适应注意力稀疏化，加速长文本LLM推理。 large language model
6 A Unified Understanding and Evaluation of Steering Methods 提出统一框架，分析并评估大语言模型中的隐空间操控方法 large language model
7 Multi-Agent Design: Optimizing Agents with Better Prompts and Topologies 提出MASS框架，通过优化提示词和拓扑结构，自动设计高效的多智能体系统。 large language model
8 Layer by Layer: Uncovering Hidden Representations in Language Models 揭示语言模型中间层表征能力，超越传统末层输出范式 large language model

🔬 支柱二：RL算法与架构 (RL & Architecture) (4 篇)

#题目一句话要点标签🔗⭐
9 Efficient Domain Adaptation of Multimodal Embeddings using Constrastive Learning 提出基于对比学习的高效多模态嵌入域自适应方法，适用于资源受限场景。 contrastive learning large language model multimodal
10 Policy Guided Tree Search for Enhanced LLM Reasoning 提出策略引导树搜索(PGTS)框架，提升LLM复杂推理能力并降低计算成本 reinforcement learning large language model chain-of-thought
11 On Teacher Hacking in Language Model Distillation 研究知识蒸馏中的“教师黑客”现象，揭示数据多样性对模型泛化的影响 reinforcement learning RLHF distillation
12 Achieving Hiding and Smart Anti-Jamming Communication: A Parallel DRL Approach against Moving Reactive Jammer 提出并行深度强化学习方法以解决移动反应干扰问题 reinforcement learning deep reinforcement learning DRL

🔬 支柱一：机器人控制 (Robot Control) (2 篇)

#题目一句话要点标签🔗⭐
13 Avoiding spurious sharpness minimization broadens applicability of SAM Functional-SAM通过避免虚假锐度最小化，扩展SAM在NLP领域的适用性 manipulation large language model
14 Neurons Speak in Ranges: Breaking Free from Discrete Neuronal Attribution 提出 NeuronLens，通过神经元激活范围解释和操控LLM，提升干预精度。 manipulation large language model

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

#题目一句话要点标签🔗⭐
15 Learning the RoPEs: Better 2D and 3D Position Encodings with STRING 提出STRING：可分离平移不变位置编码，提升2D/3D视觉Transformer性能 open-vocabulary open vocabulary large language model

⬅️ 返回 cs.LG 首页 · 🏠 返回主页