cs.CL（2026-04-20）

📊 共 43 篇论文 | 🔗 10 篇有代码

🎯 兴趣领域导航

支柱九：具身大模型 (Embodied Foundation Models) (34 🔗8) 支柱二：RL算法与架构 (RL & Architecture) (7 🔗2) 支柱六：视频提取与匹配 (Video Extraction) (1) 支柱五：交互与反应 (Interaction & Reaction) (1)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (34 篇)

#题目一句话要点标签🔗⭐
1 PDDL-Mind: Large Language Models are Capable on Belief Reasoning with Reliable State Tracking PDDL-Mind：利用大语言模型进行可靠状态追踪的信念推理 large language model chain-of-thought
2 Retrieval-Augmented Multimodal Model for Fake News Detection 提出检索增强多模态模型RAMM，解决假新闻检测中跨实例叙事一致性和领域知识缺乏问题。 large language model multimodal ✅
3 How Creative Are Large Language Models in Generating Molecules? 首个系统性研究：评估大语言模型在分子生成中的创造性表现 large language model
4 TLoRA: Task-aware Low Rank Adaptation of Large Language Models TLoRA：面向任务的大语言模型低秩自适应优化框架 large language model
5 MHSafeEval: Role-Aware Interaction-Level Evaluation of Mental Health Safety in Large Language Models 提出MHSafeEval框架，用于评估大型语言模型在心理健康咨询中的角色感知交互安全性 large language model
6 River-LLM: Large Language Model Seamless Exit Based on KV Share River-LLM：基于KV共享的大语言模型无缝退出框架，提升推理速度。 large language model
7 Omni-Embed-Audio: Leveraging Multimodal LLMs for Robust Audio-Text Retrieval 提出Omni-Embed-Audio，利用多模态LLM提升音频-文本检索的鲁棒性 multimodal
8 Multiplication in Multimodal LLMs: Computation with Text, Image, and Audio Inputs 提出多模态乘法基准测试，揭示多模态LLM在算术计算中的能力瓶颈。 multimodal
9 FLiP: Towards understanding and interpreting multimodal multilingual sentence embeddings 提出FLiP模型，用于理解和解释多模态多语言句子嵌入 multimodal ✅
10 JudgeMeNot: Personalizing Large Language Models to Emulate Judicial Reasoning in Hebrew 提出JudgeMeNot，通过个性化大语言模型模拟希伯来语司法推理 large language model
11 Employing General-Purpose and Biomedical Large Language Models with Advanced Prompt Engineering for Pharmacoepidemiologic Study Design 利用通用和生物医学大语言模型及高级Prompt工程改进药物流行病学研究设计 large language model
12 Learning to Seek Help: Dynamic Collaboration Between Small and Large Language Models 提出动态协作框架，协同大小语言模型解决多步推理问题 large language model
13 DeInfer: Efficient Parallel Inferencing for Decomposed Large Language Models 提出DeInfer，加速分解大语言模型的并行推理。 large language model
14 MM-JudgeBias: A Benchmark for Evaluating Compositional Biases in MLLM-as-a-Judge 提出MM-JudgeBias基准，用于评估多模态大语言模型作为评判者时的组合偏差。 large language model multimodal
15 Heterogeneity in Formal Linguistic Competence of Language Models: Is Data the Real Bottleneck? 通过注入少量针对性数据，显著提升小规模语言模型在特定语言现象上的理解能力。 large language model ✅
16 Forget What Matters, Keep the Rest: Selective Unlearning of Informative Tokens 提出熵引导的Token权重（ETW）方法，用于大语言模型中信息量Token的选择性遗忘，提升模型效用。 large language model
17 Multilingual Training and Evaluation Resources for Vision-Language Models 构建多语言视觉-语言模型训练与评估资源，提升非英语环境性能。 multimodal
18 Beyond Reproduction: A Paired-Task Framework for Assessing LLM Comprehension and Creativity in Literary Translation 提出配对任务框架，评估LLM在文学翻译中的理解与创造力 large language model
19 Latent Abstraction for Retrieval-Augmented Generation 提出LAnR：一种在LLM隐空间内进行检索增强生成的新框架 large language model
20 Bridging the Reasoning Gap in Vietnamese with Small Language Models via Test-Time Scaling 通过测试时缩放，利用小型语言模型弥合越南语推理差距 chain-of-thought
21 SPENCE: A Syntactic Probe for Detecting Contamination in NL2SQL Benchmarks SPENCE：一种用于检测NL2SQL基准测试集中污染的句法探针 large language model
22 Dual Alignment Between Language Model Layers and Human Sentence Processing 通过双重对齐语言模型层级与人类句子处理，提升句法复杂场景下的认知努力建模。 large language model
23 MASS-RAG: Multi-Agent Synthesis Retrieval-Augmented Generation 提出MASS-RAG，一种多智能体合成的检索增强生成方法，提升噪声环境下知识整合效果。 large language model
24 BhashaSutra: A Task-Centric Unified Survey of Indian NLP Datasets, Corpora, and Resources BhashaSutra：印度NLP数据集、语料库和资源的以任务为中心的统一综述 multimodal
25 Understanding the Prompt Sensitivity 分析LLM梯度与概率关系，揭示Prompt敏感性内在原因 large language model ✅
26 ArbGraph: Conflict-Aware Evidence Arbitration for Reliable Long-Form Retrieval-Augmented Generation ArbGraph：面向长文本RAG，提出冲突感知的证据仲裁框架，提升生成可靠性。 large language model ✅
27 HiGMem: A Hierarchical and LLM-Guided Memory System for Long-Term Conversational Agents 提出HiGMem：一种层级化和LLM引导的记忆系统，用于长期对话Agent。 large language model ✅
28 Beyond Pattern Matching: Seven Cross-Domain Techniques for Prompt Injection Detection 提出七种跨领域技术用于提示注入检测，突破传统模式匹配和微调分类器的局限。 large language model
29 Decisive: Guiding User Decisions with Optimal Preference Elicitation from Unstructured Documents Decisive：通过非结构化文档的最优偏好诱导，引导用户决策。 large language model
30 Efficient Low-Resource Language Adaptation via Multi-Source Dynamic Logit Fusion 提出TriMix，通过多源动态Logit融合实现高效的低资源语言模型适配 large language model
31 Process Reward Models Meet Planning: Generating Precise and Scalable Datasets for Step-Level Rewards 提出基于规划领域定义语言的流程奖励模型数据集生成方法，提升LLM推理能力。 large language model
32 Automatic Slide Updating with User-Defined Dynamic Templates and Natural Language Instructions 提出DynaSlide基准和SlideAgent框架，实现基于自然语言指令和用户自定义模板的幻灯片自动更新。 multimodal ✅
33 HiRAS: A Hierarchical Multi-Agent Framework for Paper-to-Code Generation and Execution 提出HiRAS层级多智能体框架，用于论文到代码生成与执行，提升实验结果复现的鲁棒性与性能。 large language model ✅
34 Do LLMs Use Cultural Knowledge Without Being Told? A Multilingual Evaluation of Implicit Pragmatic Adaptation 评估大语言模型在隐含文化情境下的语用适应能力，揭示其对文化知识的利用程度。 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🔗⭐
35 Audio-DeepThinker: Progressive Reasoning-Aware Reinforcement Learning for High-Quality Chain-of-Thought Emergence in Audio Language Models 提出Audio-DeepThinker，通过强化学习提升音频语言模型中的高质量思维链推理能力 reinforcement learning chain-of-thought
36 RePrompT: Recurrent Prompt Tuning for Integrating Structured EHR Encoders with Large Language Models RePrompT：通过循环Prompt Tuning整合结构化EHR编码器与大语言模型 predictive model large language model
37 StepPO: Step-Aligned Policy Optimization for Agentic Reinforcement Learning StepPO：面向Agentic强化学习的步对齐策略优化方法 reinforcement learning RLHF large language model
38 CodePivot: Bootstrapping Multilingual Transpilation in LLMs via Reinforcement Learning without Parallel Corpora CodePivot：利用强化学习和中间表示，在LLM中引导多语言代码转译，无需平行语料。 reinforcement learning large language model ✅
39 Aligning Language Models for Lyric-to-Melody Generation with Rule-Based Musical Constraints 提出基于规则约束对齐的语言模型，用于歌词到旋律生成，显著提升音乐性。 DPO direct preference optimization large language model ✅
40 PRISMA: Preference-Reinforced Self-Training Approach for Interpretable Emotionally Intelligent Negotiation Dialogues PRISMA：一种基于偏好强化自训练的可解释情感智能协商对话系统 DPO direct preference optimization chain-of-thought
41 Reverse Constitutional AI: A Framework for Controllable Toxic Data Generation via Probability-Clamped RLAIF 提出R-CAI框架，通过概率钳制的RLAIF实现可控的有害数据生成。 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 Culture-Aware Humorous Captioning: Multimodal Humor Generation across Cultural Contexts 提出文化感知幽默字幕生成以解决多模态幽默生成的文化适应性问题 HuMoR large language model multimodal

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

#题目一句话要点标签🔗⭐
43 FUSE: Ensembling Verifiers with Zero Labeled Data FUSE：一种无需标注数据的验证器集成方法，提升LLM输出质量 IMoS large language model

⬅️ 返回 cs.CL 首页 · 🏠 返回主页