cs.CL（2025-09-29）

📊 共 61 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) (44 🔗9) 支柱二：RL算法与架构 (RL & Architecture) (13 🔗1) 支柱八：物理动画 (Physics-based Animation) (1) 支柱七：动作重定向 (Motion Retargeting) (1) 支柱一：机器人控制 (Robot Control) (1) 支柱六：视频提取与匹配 (Video Extraction) (1)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (44 篇)

#题目一句话要点标签🔗⭐
1 Multimodal Large Language Models Meet Multimodal Emotion Recognition and Reasoning: A Survey 综述多模态大语言模型在情感识别与推理中的应用与挑战 large language model multimodal ✅
2 Metaphor identification using large language models: A comparison of RAG, prompt engineering, and fine-tuning 利用大型语言模型进行隐喻识别：比较RAG、提示工程和微调方法 large language model chain-of-thought
3 Dual Mechanisms of Value Expression: Intrinsic vs. Prompted Values in Large Language Models 研究揭示大语言模型中内在价值观与提示诱导价值观的双重机制 large language model instruction following
4 Prompt and Parameter Co-Optimization for Large Language Models 提出MetaTuner框架，联合优化Prompt和参数以提升大语言模型性能 large language model
5 Can Large Language Models Express Uncertainty Like Human? 提出语言置信度方法，提升大语言模型不确定性表达能力 large language model
6 Beyond Overall Accuracy: A Psychometric Deep Dive into the Topic-Specific Medical Capabilities of 80 Large Language Models 提出MedIRT框架，利用项目反应理论评估LLM的医学能力，揭示模型专长与缺陷。 large language model
7 Towards Structured Knowledge: Advancing Triple Extraction from Regional Trade Agreements using Large Language Models 利用大型语言模型从区域贸易协定中提取结构化知识，构建贸易知识图谱。 large language model
8 Learning to Parallel: Accelerating Diffusion Large Language Models via Learnable Parallel Decoding 提出Learn2PD以解决大语言模型推理速度瓶颈问题 large language model
9 Pretraining Large Language Models with NVFP4 提出NVFP4训练方法，实现4比特精度下大语言模型的稳定高效预训练。 large language model
10 GateMABSA: Aspect-Image Gated Fusion for Multimodal Aspect-based Sentiment Analysis 提出GateMABSA模型，通过门控多模态融合解决多模态情感分析中噪声过滤和跨模态对齐问题。 multimodal
11 Understanding the Dilemma of Unlearning for Large Language Models 提出unPact框架，揭示大语言模型不可靠的知识遗忘现象与灾难性遗忘困境 large language model
12 Sanitize Your Responses: Mitigating Privacy Leakage in Large Language Models 提出Self-Sanitize框架，缓解大语言模型中的隐私泄露问题 large language model ✅
13 CDT: A Comprehensive Capability Framework for Large Language Models Across Cognition, Domain, and Task 提出CDT框架，从认知、领域和任务三维度全面评估大语言模型能力。 large language model ✅
14 AlignX: Advancing Multilingual Large Language Models with Multilingual Representation Alignment AlignX：通过多语言表示对齐提升多语言大语言模型性能 large language model
15 DiffuGuard: How Intrinsic Safety is Lost and Found in Diffusion Large Language Models DiffuGuard：揭示并修复扩散大语言模型中固有的安全漏洞 large language model ✅
16 MobileLLM-R1: Exploring the Limits of Sub-Billion Language Model Reasoners with Open Training Recipes MobileLLM-R1：通过开放训练方案探索十亿参数以下语言模型推理能力的极限 large language model chain-of-thought
17 InfLLM-V2: Dense-Sparse Switchable Attention for Seamless Short-to-Long Adaptation 提出InfLLM-V2：一种稠密-稀疏可切换注意力机制，实现模型从短序列到长序列的无缝适应。 large language model chain-of-thought ✅
18 AdaThink-Med: Medical Adaptive Thinking with Uncertainty-Guided Length Calibration AdaThink-Med：提出不确定性引导长度校准的医学自适应思考框架 large language model chain-of-thought
19 AceSearcher: Bootstrapping Reasoning and Search for LLMs via Reinforced Self-Play AceSearcher：通过强化自博弈引导LLM进行推理和搜索，提升复杂推理任务性能。 large language model ✅
20 SPECTRA: Revealing the Full Spectrum of User Preferences via Distributional LLM Inference SPECTRA：通过分布式的LLM推理揭示用户偏好的全谱，解决推荐系统中的长尾偏好问题。 large language model
21 Task Vectors, Learned Not Extracted: Performance Gains and Mechanistic Insight 提出可学习任务向量(LTV)，提升ICL性能并提供机制性理解 large language model
22 Localizing Task Recognition and Task Learning in In-Context Learning via Attention Head Analysis 通过注意力头分析，在上下文学习中定位任务识别和任务学习 large language model
23 Calibrating Verbalized Confidence with Self-Generated Distractors 提出DINCO方法，通过自生成干扰项校准LLM的置信度，提升可靠性。 large language model
24 Not Wrong, But Untrue: LLM Overconfidence in Document-Based Queries LLM在文档问答中过度自信：揭示新闻场景下的幻觉问题与溯源挑战 large language model
25 The Rise of AfricaNLP: Contributions, Contributors, and Community Impact (2005-2025) 追踪非洲NLP发展：贡献分析、参与者识别与社区影响评估 large language model
26 Fingerprinting LLMs via Prompt Injection LLMPrint：利用Prompt注入为LLM构建鲁棒指纹，实现模型溯源 large language model
27 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
28 From Internal Representations to Text Quality: A Geometric Approach to LLM Evaluation 利用LLM内部表征的几何特性评估文本质量，实现无参考文本质量评估 large language model
29 Investigating Language and Retrieval Bias in Multilingual Previously Fact-Checked Claim Detection 研究多语言预训练模型在事实核查中的语言和检索偏差 large language model
30 Learning from Convenience Samples: A Case Study on Fine-Tuning LLMs for Survey Non-response in the German Longitudinal Election Study 利用便利样本微调LLM，解决德国选举研究中调查无应答问题 large language model
31 Hyperdimensional Probe: Decoding LLM Representations via Vector Symbolic Architectures 提出超维探针，通过向量符号架构解码大型语言模型表征 large language model
32 How Well Do LLMs Imitate Human Writing Style? 提出免训练风格模仿分析框架，评估LLM模仿人类写作风格的能力 large language model
33 BOE-XSUM: Extreme Summarization in Clear Language of Spanish Legal Decrees and Notifications BOE-XSUM：发布西班牙法律公文的明晰语言极端摘要数据集，并验证微调LLM的有效性 large language model
34 Expanding Computation Spaces of LLMs at Inference Time 提出推理时LLM计算空间扩展方法，提升开放域问答和数学任务性能 chain-of-thought
35 SemShareKV: Efficient KVCache Sharing for Semantically Similar Prompts via Token-Level LSH Matching SemShareKV：通过Token级LSH匹配为语义相似Prompt高效共享KVCache large language model
36 Hallucination is Inevitable for LLMs with the Open World Assumption 重新定义LLM幻觉：开放世界假设下，幻觉是大型语言模型泛化问题的必然结果 large language model
37 ProxyAttn: Guided Sparse Attention via Representative Heads ProxyAttn：通过代表性注意力头引导的稀疏注意力机制，加速长文本处理。 large language model ✅
38 Think Twice, Generate Once: Safeguarding by Progressive Self-Reflection 提出渐进式自反思（PSR）方法，提升LLM在生成任务中的安全性。 large language model
39 MemGen: Weaving Generative Latent Memory for Self-Evolving Agents MemGen：为自进化Agent构建生成式潜在记忆，提升认知能力 large language model
40 Bias Mitigation or Cultural Commonsense? Evaluating LLMs with a Japanese Dataset 提出SOBACO日语基准，评估LLM偏见缓解对文化常识的影响 large language model
41 HarmMetric Eval: Benchmarking Metrics and Judges for LLM Harmfulness Assessment 提出HarmMetric Eval，用于全面评估LLM有害性评估指标与判别器的质量。 large language model ✅
42 MAS$^2$: Self-Generative, Self-Configuring, Self-Rectifying Multi-Agent Systems 提出MAS$^2$，一种自生成、自配置、自校正的多智能体系统，提升复杂任务性能。 large language model ✅
43 How Training Data Shapes the Use of Parametric and In-Context Knowledge in Language Models 揭示训练数据特性如何影响语言模型参数知识和上下文知识的利用 large language model
44 SimuHome: A Temporal- and Environment-Aware Benchmark for Smart Home LLM Agents SimuHome：面向智能家居LLM代理的时间与环境感知基准测试 large language model

🔬 支柱二：RL算法与架构 (RL & Architecture) (13 篇)

#题目一句话要点标签🔗⭐
45 GRPO-MA: Multi-Answer Generation in GRPO for Stable and Efficient Chain-of-Thought Training GRPO-MA：通过多答案生成提升GRPO在CoT训练中的稳定性和效率 reinforcement learning large language model multimodal
46 RFG: Test-Time Scaling for Diffusion Large Language Model Reasoning with Reward-Free Guidance 提出RFG：一种免奖励引导的扩散大语言模型推理测试时缩放方法 reinforcement learning large language model
47 SeaPO: Strategic Error Amplification for Robust Preference Optimization of Large Language Models SeaPO：通过策略性误差放大增强大语言模型偏好优化的鲁棒性 preference learning large language model
48 Let LLMs Speak Embedding Languages: Generative Text Embeddings via Iterative Contrastive Refinement 提出GIRCSE，利用生成式LLM迭代优化文本嵌入，提升语义表征能力 representation learning large language model instruction following
49 Aligning Multilingual Reasoning with Verifiable Semantics from a High-Resource Expert Model 提出PB-RLSVR框架，利用高资源专家模型提升多语言LLM的推理能力。 reinforcement learning PPO large language model
50 InfoAgent: Advancing Autonomous Information-Seeking Agents InfoAgent：通过数据合成和自建搜索工具提升自主信息搜寻能力 reinforcement learning large language model
51 Towards Trustworthy Lexical Simplification: Exploring Safety and Efficiency with Small LLMs 提出基于小型LLM的安全高效词汇简化框架，解决隐私和资源约束下的应用难题。 distillation large language model
52 Circuit Distillation 提出电路蒸馏方法，通过对齐模型内部表征实现算法能力迁移 distillation
53 Socratic-Zero : Bootstrapping Reasoning via Data-Free Agent Co-evolution Socratic-Zero：通过无数据Agent协同进化引导LLM推理能力 distillation large language model
54 AdaDetectGPT: Adaptive Detection of LLM-Generated Text with Statistical Guarantees AdaDetectGPT：利用统计保证自适应检测LLM生成文本 Mamba large language model ✅
55 Alternatives To Next Token Prediction In Text Generation -- A Survey 综述：探索文本生成中下一词预测的替代方案，应对LLM的固有缺陷。 flow matching large language model
56 Reinforcement Mid-Training 提出强化中期训练（RMT）框架，提升大语言模型性能并加速训练。 reinforcement learning large language model
57 Beyond Repetition: Text Simplification and Curriculum Learning for Data-Constrained Pretraining 针对数据受限的预训练，提出基于文本简化和课程学习的优化方法。 curriculum learning

🔬 支柱八：物理动画 (Physics-based Animation) (1 篇)

#题目一句话要点标签🔗⭐
58 Evaluating Spatiotemporal Consistency in Automatically Generated Sewing Instructions 提出一种基于树结构的自动评估指标，用于评估LLM生成的缝纫指令的时空一致性。 spatiotemporal

🔬 支柱七：动作重定向 (Motion Retargeting) (1 篇)

#题目一句话要点标签🔗⭐
59 LatentEvolve: Self-Evolving Test-Time Scaling in Latent Space 提出LatentEvolve，通过潜空间自进化测试时缩放提升LLM推理能力。 latent optimization large language model

🔬 支柱一：机器人控制 (Robot Control) (1 篇)

#题目一句话要点标签🔗⭐
60 EasySteer: A Unified Framework for High-Performance and Extensible LLM Steering EasySteer：高性能、可扩展的LLM引导统一框架，加速可控语言模型部署 manipulation 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🔗⭐
61 Probing the Limits of Stylistic Alignment in Vision-Language Models 研究视觉-语言模型在风格对齐上的极限，探索数据效率与性能饱和点 HuMoR

⬅️ 返回 cs.CL 首页 · 🏠 返回主页