cs.CL（2026-02-28）

📊 共 43 篇论文

🎯 兴趣领域导航

支柱九：具身大模型 (Embodied Foundation Models) (31) 支柱二：RL算法与架构 (RL & Architecture) (10) 支柱六：视频提取与匹配 (Video Extraction) (1) 支柱一：机器人控制 (Robot Control) (1)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (31 篇)

#题目一句话要点标签🔗⭐
1 TCM-DiffRAG: Personalized Syndrome Differentiation Reasoning Method for Traditional Chinese Medicine based on Knowledge Graph and Chain of Thought TCM-DiffRAG：基于知识图谱和思维链的中医个性化辨证论治方法 large language model chain-of-thought
2 A Mixture-of-Experts Model for Multimodal Emotion Recognition in Conversations 提出MiSTER-E模型，通过混合专家机制解决对话情感识别中的多模态融合问题。 large language model multimodal
3 Parallel Continuous Chain-of-Thought with Jacobi Iteration 提出基于Jacobi迭代的并行连续思维链PCCoT，加速LLM推理。 large language model chain-of-thought
4 Modality Collapse as Mismatched Decoding: Information-Theoretic Limits of Multimodal LLMs 提出多模态LLM解码不匹配问题以提升信息提取能力 multimodal
5 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，加速LLM推理。 large language model
6 Under the Influence: Quantifying Persuasion and Vigilance in Large Language Models 量化大语言模型中的说服力与警惕性，揭示AI安全新视角 large language model
7 Tokenization, Fusion and Decoupling: Bridging the Granularity Mismatch Between Large Language Models and Knowledge Graphs 提出KGT框架以解决大语言模型与知识图谱间的粒度不匹配问题 large language model
8 Probing for Knowledge Attribution in Large Language Models 提出AttriWiki自监督数据管道，用于探究大语言模型知识归属问题 large language model
9 Mind the Gap in Cultural Alignment: Task-Aware Culture Management for Large Language Models 提出CultureManager，解决大语言模型中任务相关的文化对齐问题 large language model
10 When Large Multimodal Models Confront Evolving Knowledge: Challenges and Explorations 提出MMEVOKE基准，探索大模型在多模态演进知识注入中的挑战与方法。 multimodal
11 DeVisE: Behavioral Testing of Medical Large Language Models DeVisE：通过行为测试评估医学大型语言模型在临床推理中的稳健性 large language model
12 Large Language Models are Algorithmically Blind 揭示大语言模型在算法推理上的局限性：算法盲区 large language model
13 Not All Attention is Needed: Parameter and Computation Efficient Transfer Learning for Multi-modal Large Language Models 提出高效注意力跳跃（EAS）方法，加速多模态大语言模型推理并保持参数效率。 large language model
14 Natural Language Declarative Prompting (NLD-P): A Modular Governance Method for Prompt Design Under Model Drift 提出自然语言声明式提示（NLD-P）方法，应对大语言模型漂移下的提示工程挑战。 large language model instruction following
15 Imagination Helps Visual Reasoning, But Not Yet in Latent Space 质疑隐空间推理有效性，提出显式文本想象方法CapImagine提升视觉推理性能 large language model multimodal
16 SQaLe: A Large Text-to-SQL Corpus Grounded in Real Schemas 提出SQaLe：一个基于真实Schema的大规模Text-to-SQL数据集，提升模型泛化能力。 large language model
17 Causality $\neq$ Invariance: Function and Concept Vectors in LLMs 揭示LLM中函数向量非不变性：提出概念向量以提升跨领域泛化能力 large language model
18 TARAZ: Persian Short-Answer Question Benchmark for Cultural Evaluation of Language Models 提出TARAZ：波斯语短答案题基准，用于评估语言模型的文化理解能力 large language model
19 Assessing Deanonymization Risks with Stylometry-Assisted LLM Agent 提出SALA框架，利用文体学特征辅助LLM代理评估和降低文本数据中的去匿名化风险。 large language model
20 Generative Value Conflicts Reveal LLM Priorities ConflictScope：揭示LLM在价值冲突下的优先级偏好，并提出系统提示对齐方法。 large language model
21 Scaling In, Not Up? Testing Thick Citation Context Analysis with GPT-5 and Fragile Prompts 通过深入文本分析而非类型标签扩展，评估GPT-5在引文情境分析中的能力及脆弱提示的影响。 large language model
22 Ruyi2 Technical Report Ruyi2：基于Familial Model的自适应深度计算加速方案 large language model
23 Towards Simulating Social Media Users with LLMs: Evaluating the Operational Validity of Conditioned Comment Prediction 提出条件化评论预测CCP任务，评估LLM模拟社交媒体用户行为的有效性 large language model
24 Test-Time Scaling with Diffusion Language Models via Reward-Guided Stitching 提出基于奖励引导拼接的扩散语言模型测试时缩放方法，提升复杂推理任务性能。 large language model
25 Bridging Latent Reasoning and Target-Language Generation via Retrieval-Transition Heads 通过检索-转换头桥接潜在推理和目标语言生成 chain-of-thought
26 Rejection Mixing: Fast Semantic Propagation of Mask Tokens for Efficient DLLM Inference 提出ReMix，通过连续空间语义传播加速DLLM推理，解决组合矛盾问题。 large language model
27 MTRAG-UN: A Benchmark for Open Challenges in Multi-Turn RAG Conversations MTRAG-UN：用于多轮RAG对话开放挑战的基准数据集 large language model
28 Can LLMs Simulate Human Behavioral Variability? A Case Study in the Phonemic Fluency Task 评估大型语言模型在音素流畅性任务中模拟人类行为变异性的能力 large language model
29 A Third Paradigm for LLM Evaluation: Dialogue Game-Based Evaluation using clembench 提出clembench，一种基于对话游戏的LLM评估框架，易于扩展和复用。 large language model
30 UPDESH: Synthesizing Grounded Instruction Tuning Data for 13 Indic Languages UPDESH：合成13种印度语言的指令微调数据，提升多语言AI性能 instruction following
31 Fine-tuning Done Right in Model Editing 重塑微调：提出LocFT-BF，显著提升模型编辑性能并扩展至更大规模模型 large language model

🔬 支柱二：RL算法与架构 (RL & Architecture) (10 篇)

#题目一句话要点标签🔗⭐
32 Knowledge Distillation with Structured Chain-of-Thought for Text-to-SQL 提出Struct-SQL框架，利用结构化CoT蒸馏提升Text-to-SQL小模型的性能。 distillation large language model chain-of-thought
33 Evaluating the Diversity and Quality of LLM Generated Content 提出有效语义多样性评估框架，揭示偏好调整模型在生成高质量多样化内容方面的优势。 reinforcement learning PPO RLHF
34 Search-P1: Path-Centric Reward Shaping for Stable and Efficient Agentic RAG Training 提出Search-P1框架，通过路径中心奖励塑造提升Agentic RAG训练的稳定性和效率。 reward shaping large language model
35 Fine-Tuning Without Forgetting In-Context Learning: A Theoretical Analysis of Linear Attention Models 提出线性注意力模型的微调方法以解决上下文学习退化问题 linear attention large language model
36 Supervised Reinforcement Learning: From Expert Trajectories to Step-wise Reasoning 提出监督强化学习(SRL)，解决小模型多步推理难题。 reinforcement learning large language model
37 Decoder-based Sense Knowledge Distillation 提出基于解码器的语义知识蒸馏框架DSKD，提升生成式LLM的语义知识。 distillation large language model
38 PARL: Prompt-based Agents for Reinforcement Learning PARL：提出一种基于提示的大语言模型强化学习智能体，无需微调。 reinforcement learning large language model
39 Revisiting Self-Play Preference Optimization: On the Role of Prompt Difficulty 研究提示词难度对自博弈偏好优化性能的影响，并提出优化策略。 DPO direct preference optimization large language model
40 Document Reconstruction Unlocks Scalable Long-Context RLVR 提出基于文档重构的无监督RLVR方法，提升LLM长文本处理能力 reinforcement learning reward design large language model
41 Towards Better RL Training Data Utilization via Second-Order Rollout 提出二阶Rollout，提升RL训练数据利用率，增强LLM生成与评价能力 reinforcement learning large language model

🔬 支柱六：视频提取与匹配 (Video Extraction) (1 篇)

#题目一句话要点标签🔗⭐
42 Detecting Hate and Inflammatory Content in Bengali Memes: A New Multimodal Dataset and Co-Attention Framework 提出MCFM模型和Bn-HIB数据集，用于检测孟加拉语Meme中的仇恨和煽动性内容 HuMoR multimodal

🔬 支柱一：机器人控制 (Robot Control) (1 篇)

#题目一句话要点标签🔗⭐
43 Replacing Multi-Step Assembly of Data Preparation Pipelines with One-Step LLM Pipeline Generation for Table QA 提出Operation-R1框架，用单步LLM生成数据准备流水线解决表格问答问题。 manipulation reinforcement learning large language model

⬅️ 返回 cs.CL 首页 · 🏠 返回主页