cs.CL（2025-10-30）

📊 共 36 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) (27 🔗2) 支柱二：RL算法与架构 (RL & Architecture) (7 🔗1) 支柱五：交互与反应 (Interaction & Reaction) (1) 支柱一：机器人控制 (Robot Control) (1)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (27 篇)

#题目一句话要点标签🔗⭐
1 Inference-Cost-Aware Dynamic Tree Construction for Efficient Inference in Large Language Models 提出CAST：一种推理成本感知的动态树构建方法，提升大语言模型推理效率 large language model
2 The Structure of Relation Decoding Linear Operators in Large Language Models 揭示大语言模型关系解码线性算子的结构，发现其主要基于语义属性而非特定关系。 large language model
3 Evontree: Ontology Rule-Guided Self-Evolution of Large Language Models Evontree：利用本体规则引导大语言模型自进化，提升领域知识 large language model
4 Bayesian Network Fusion of Large Language Models for Sentiment Analysis 提出基于贝叶斯网络的大语言模型融合框架，提升情感分析性能。 large language model
5 Encoder-Decoder or Decoder-Only? Revisiting Encoder-Decoder Large Language Model 重新审视Encoder-Decoder大语言模型，探索其在效率和性能上的潜力 large language model
6 A Multi-agent Large Language Model Framework to Automatically Assess Performance of a Clinical AI Triage Tool 提出多Agent LLM框架，自动评估临床AI分诊工具的性能 large language model
7 1+1>2: A Synergistic Sparse and Low-Rank Compression Method for Large Language Models 提出协同稀疏与低秩压缩方法SSLC，高效压缩大型语言模型。 large language model
8 OmniEduBench: A Comprehensive Chinese Benchmark for Evaluating Large Language Models in Education 提出OmniEduBench，用于全面评估中文教育领域大语言模型能力 large language model
9 RCScore: Quantifying Response Consistency in Large Language Models RCScore：量化大语言模型对指令形式的响应一致性，评估模型鲁棒性 large language model
10 MisSynth: Improving MISSCI Logical Fallacies Classification with Synthetic Data MisSynth：利用合成数据提升MISSCI谬误逻辑分类性能 large language model ✅
11 Unravelling the Mechanisms of Manipulating Numbers in Language Models 揭示语言模型中数字处理机制，探究其误差根源与精度下限 large language model
12 PVMark: Enabling Public Verifiability for LLM Watermarking Schemes PVMark：一种支持LLM水印方案公开可验证性的框架 large language model
13 Detecting Data Contamination in LLMs via In-Context Learning 提出CoDeC，通过上下文学习检测LLM中的数据污染 large language model
14 Chopping Trees: Semantic Similarity Based Dynamic Pruning for Tree-of-Thought Reasoning 提出基于语义相似性的动态剪枝方法，加速思维树推理。 large language model ✅
15 The Geometry of Dialogue: Graphing Language Models to Reveal Synergistic Teams for Multi-Agent Collaboration 提出基于语言模型图的协同团队构建方法，解决多智能体LLM协作中的团队优化问题。 large language model
16 Towards Global Retrieval Augmented Generation: A Benchmark for Corpus-Level Reasoning 提出GlobalRAG框架，解决现有RAG方法在语料库级别推理任务中的不足。 large language model
17 VISTA Score: Verification In Sequential Turn-based Assessment VISTA：提出一种用于评估对话系统中事实性幻觉的序列轮次验证框架 large language model
18 Kad: A Framework for Proxy-based Test-time Alignment with Knapsack Approximation Deferral Kad框架：基于代理模型的测试时对齐，利用背包近似延迟解决LLM对齐计算成本高昂问题。 large language model
19 Semantically-Aware LLM Agent to Enhance Privacy in Conversational AI Services 提出LOPSIDED框架，增强会话AI中LLM的隐私保护能力 large language model
20 SlideAgent: Hierarchical Agentic Framework for Multi-Page Visual Document Understanding 提出SlideAgent，用于多页视觉文档理解的分层Agent框架。 large language model
21 On the Role of Context for Discourse Relation Classification in Scientific Writing 研究科学写作中篇章关系分类任务，探讨上下文信息对提升性能的作用 large language model
22 Do LLMs Signal When They're Right? Evidence from Neuron Agreement 提出神经元一致性解码(NAD)，利用LLM内部神经元信号提升无标签集成解码效果。 large language model
23 Language Models Are Borrowing-Blind: A Multilingual Evaluation of Loanword Identification across 10 Languages 大型语言模型在跨语种外来语识别任务中表现不佳，揭示其“借用盲区”问题 large language model
24 Pragmatic Theories Enhance Understanding of Implied Meanings in LLMs 利用语用理论提示提升LLM对隐含意义的理解能力 chain-of-thought
25 Similarity-Distance-Magnitude Language Models 提出基于相似度-距离-幅度(SDM)激活的语言模型，提升指令跟随任务的统计效率。 instruction following
26 On the Influence of Discourse Relations in Persuasive Texts 利用大型语言模型分析说服文本中论述关系对说服技巧的影响 large language model
27 QCoder Benchmark: Bridging Language Generation and Quantum Hardware through Simulator-Based Feedback 提出QCoder Benchmark，通过模拟器反馈评估LLM在量子编程中的代码生成能力 large language model

🔬 支柱二：RL算法与架构 (RL & Architecture) (7 篇)

#题目一句话要点标签🔗⭐
28 Reasoning Up the Instruction Ladder for Controllable Language Models 提出VerIH数据集和强化学习方法，提升LLM指令层级推理能力和安全性 reinforcement learning large language model instruction following
29 From Amateur to Master: Infusing Knowledge into LLMs via Automated Curriculum Learning ACER：通过自动化课程学习将知识注入大型语言模型，提升领域专业性。 curriculum learning large language model
30 Understanding and Enhancing Mamba-Transformer Hybrids for Memory Recall and Language Modeling 分析Mamba-Transformer混合模型，提出数据增强方法提升记忆回溯和语言建模能力。 Mamba SSM state space model
31 Evaluating Perspectival Biases in Cross-Modal Retrieval 提出3XCM基准，评估跨模态检索中由语言和文化差异引起的视角偏差。 representation learning multimodal language conditioned
32 Reasoning Path Divergence: A New Metric and Curation Strategy to Unlock LLM Diverse Thinking 提出推理路径差异度量与数据筛选策略，提升LLM推理多样性与性能 reinforcement learning large language model chain-of-thought ✅
33 MossNet: Mixture of State-Space Experts is a Multi-Head Attention 提出MossNet：一种混合状态空间专家模型，模拟多头注意力机制，提升LLM性能。 SSM large language model
34 Kimi Linear: An Expressive, Efficient Attention Architecture Kimi Linear：一种高效且富有表现力的线性注意力架构，性能超越传统全注意力。 reinforcement learning linear attention

🔬 支柱五：交互与反应 (Interaction & Reaction) (1 篇)

#题目一句话要点标签🔗⭐
35 AMO-Bench: Large Language Models Still Struggle in High School Math Competitions 提出AMO-Bench，用于评估大语言模型在奥林匹克级别数学问题上的推理能力。 IMoS large language model

🔬 支柱一：机器人控制 (Robot Control) (1 篇)

#题目一句话要点标签🔗⭐
36 Value Drifts: Tracing Value Alignment During LLM Post-Training 揭示LLM后训练阶段价值观漂移，探究价值观对齐的学习动态过程 manipulation

⬅️ 返回 cs.CL 首页 · 🏠 返回主页