cs.CL（2026-02-03）

📊 共 34 篇论文 | 🔗 5 篇有代码

🎯 兴趣领域导航

支柱九：具身大模型 (Embodied Foundation Models) (22 🔗3) 支柱二：RL算法与架构 (RL & Architecture) (11 🔗2) 支柱一：机器人控制 (Robot Control) (1)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (22 篇)

#题目一句话要点标签🔗⭐
1 Can Large Language Models Generalize Procedures Across Representations? 提出两阶段数据课程学习方法，提升LLM在代码、图和自然语言表示之间的程序泛化能力 large language model
2 Cognitively Diverse Multiple-Choice Question Generation: A Hybrid Multi-Agent Framework with Large Language Models ReQUESTA：一种混合多智能体框架，利用大型语言模型生成认知多样性的多项选择题 large language model
3 Assessing the Impact of Typological Features on Multilingual Machine Translation in the Age of Large Language Models 研究表明目标语言类型学特征显著影响大语言模型多语言翻译质量 large language model
4 ChemPro: A Progressive Chemistry Benchmark for Large Language Models ChemPro：面向大语言模型的渐进式化学能力评测基准 large language model
5 The Mask of Civility: Benchmarking Chinese Mock Politeness Comprehension in Large Language Models 构建中文嘲讽礼貌数据集，评测大型语言模型在礼貌理解上的能力。 large language model
6 Instruction Anchors: Dissecting the Causal Dynamics of Modality Arbitration 提出指令锚点理论，揭示多模态大语言模型中模态仲裁的因果机制。 large language model multimodal
7 SEAD: Self-Evolving Agent for Multi-Turn Service Dialogue 提出SEAD自进化Agent，解决服务对话中数据匮乏和用户行为模拟难题。 large language model foundation model ✅
8 POP: Prefill-Only Pruning for Efficient Large Model Inference 提出POP：一种预填充阶段剪枝方法，提升大模型推理效率 large language model
9 Token Sparse Attention: Efficient Long-Context Inference with Interleaved Token Selection 提出Token Sparse Attention，通过交错Token选择加速长文本推理。 large language model
10 LatentMem: Customizing Latent Memory for Multi-Agent Systems LatentMem：为多智能体系统定制角色感知的紧凑型隐空间记忆 large language model
11 Accelerating Scientific Research with Gemini: Case Studies and Common Techniques 利用Gemini加速科学研究：案例分析与通用技术 large language model
12 Use Graph When It Needs: Efficiently and Adaptively Integrating Retrieval-Augmented Generation with Graphs 提出EA-GraphRAG，通过语法感知复杂度分析自适应融合RAG与图增强RAG，提升知识密集型任务性能。 large language model
13 ATACompressor: Adaptive Task-Aware Compression for Efficient Long-Context Processing in LLMs 提出ATACompressor，通过自适应任务感知压缩提升LLM长文本处理效率。 large language model
14 Task--Specificity Score: Measuring How Much Instructions Really Matter for Supervision 提出任务特异性评分(TSS)以衡量指令对LLM监督的重要性，提升小样本学习性能。 large language model
15 Where Norms and References Collide: Evaluating LLMs on Normative Reasoning 提出SNIC基准测试，评估LLM在情境规范推理中的能力，揭示其在规范理解上的局限性。 large language model
16 They Said Memes Were Harmless-We Found the Ones That Hurt: Decoding Jokes, Symbols, and Cultural References 提出CROSS-ALIGN+框架，解决基于Meme的社交恶意滥用检测中文化盲区和可解释性问题。 multimodal
17 Beyond Tokens: Semantic-Aware Speculative Decoding for Efficient Inference by Probing Internal States 提出SemanticSpec，通过语义感知的推测解码加速大型语言模型的推理。 large language model
18 SWE-World: Building Software Engineering Agents in Docker-Free Environments 提出SWE-World，一个无Docker环境的软件工程Agent训练框架，提升代码修改任务性能。 large language model ✅
19 MIRROR: A Multi-Agent Framework with Iterative Adaptive Revision and Hierarchical Retrieval for Optimization Modeling in Operations Research MIRROR：面向运筹优化建模的迭代自适应修正与分层检索多Agent框架 large language model
20 FASA: Frequency-aware Sparse Attention FASA：提出频率感知稀疏注意力机制，解决长文本LLM的KV缓存瓶颈。 large language model
21 AERO: Autonomous Evolutionary Reasoning Optimization via Endogenous Dual-Loop Feedback AERO：通过内生双环反馈实现自主进化推理优化 large language model ✅
22 SAES-SVD: Self-Adaptive Suppression of Accumulated and Local Errors for SVD-based LLM Compression SAES-SVD：自适应抑制累积误差的SVD压缩LLM框架 large language model

🔬 支柱二：RL算法与架构 (RL & Architecture) (11 篇)

#题目一句话要点标签🔗⭐
23 Towards Distillation-Resistant Large Language Models: An Information-Theoretic Perspective 提出基于信息论的抗蒸馏大语言模型方法，防御Logit蒸馏攻击。 distillation large language model
24 CPMobius: Iterative Coach-Player Reasoning for Data-Free Reinforcement Learning 提出CPMobius，一种用于数据自由强化学习的迭代教练-队员推理框架 reinforcement learning large language model
25 TRE: Encouraging Exploration in the Trust Region 提出信任域熵(TRE)方法，解决LLM中探索失效问题，提升数学推理、组合搜索和偏好对齐任务性能。 reinforcement learning PPO large language model ✅
26 Learning to Reason Faithfully through Step-Level Faithfulness Maximization 提出FaithRL框架，通过最大化步骤级忠实度提升LLM推理能力，减少幻觉。 reinforcement learning reward design large language model ✅
27 Verified Critical Step Optimization for LLM Agents 提出关键步骤优化(CSO)方法，提升LLM Agent在复杂任务中的表现 preference learning DPO large language model
28 $V_0$: A Generalist Value Model for Any Policy at State Zero 提出V0通用价值模型，无需更新参数即可评估任意策略在初始状态的性能，用于LLM训练和部署。 PPO large language model
29 ForesightKV: Optimizing KV Cache Eviction for Reasoning Models by Learning Long-Term Contribution ForesightKV：通过学习长期贡献优化推理模型KV缓存淘汰 reinforcement learning large language model
30 ACL: Aligned Contrastive Learning Improves BERT and Multi-exit BERT Fine-tuning 提出对齐对比学习(ACL)框架，提升BERT和多出口BERT微调性能 contrastive learning
31 One Model, All Roles: Multi-Turn, Multi-Agent Self-Play Reinforcement Learning for Conversational Social Intelligence 提出OMAR：基于多智能体自博弈强化学习的通用对话社交智能模型 reinforcement learning
32 Test-time Recursive Thinking: Self-Improvement without External Feedback 提出测试时递归思考(TRT)，实现大语言模型无需额外训练的自我提升。 reinforcement learning large language model
33 ReMiT: RL-Guided Mid-Training for Iterative LLM Evolution ReMiT：强化学习引导的LLM中期训练，实现迭代式模型进化 reinforcement learning large language model

🔬 支柱一：机器人控制 (Robot Control) (1 篇)

#题目一句话要点标签🔗⭐
34 Controlling Output Rankings in Generative Engines for LLM-based Search 提出CORE方法，通过优化检索内容控制LLM搜索的输出排序，提升小商家产品曝光 manipulation large language model

⬅️ 返回 cs.CL 首页 · 🏠 返回主页