cs.CL（2025-10-06）

📊 共 38 篇论文 | 🔗 6 篇有代码

🎯 兴趣领域导航

支柱九：具身大模型 (Embodied Foundation Models) (26 🔗4) 支柱二：RL算法与架构 (RL & Architecture) (11 🔗1) 支柱一：机器人控制 (Robot Control) (1 🔗1)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (26 篇)

#题目一句话要点标签🔗⭐
1 Resource-Efficient Fine-Tuning of LLaMA-3.2-3B for Medical Chain-of-Thought Reasoning 提出一种资源高效的LLaMA-3.2-3B微调方法，用于提升医疗领域的CoT推理能力 large language model chain-of-thought
2 Finish First, Perfect Later: Test-Time Token-Level Cross-Validation for Diffusion Large Language Models 提出Tolerator，通过Token级交叉验证优化扩散大语言模型的解码策略。 large language model
3 Imperceptible Jailbreaking against Large Language Models 提出基于不可见Unicode变异选择器的LLM越狱攻击方法 large language model ✅
4 Reproducibility Study of "XRec: Large Language Models for Explainable Recommendation" 复现XRec：基于大语言模型的可解释推荐框架，并探索MoE嵌入的影响 large language model ✅
5 FocusMed: A Large Language Model-based Framework for Enhancing Medical Question Summarization with Focus Identification FocusMed：基于大语言模型的医疗问答摘要框架，增强焦点识别能力 large language model ✅
6 A Lightweight Large Language Model-Based Multi-Agent System for 2D Frame Structural Analysis 提出基于轻量级大语言模型的多Agent系统，用于2D框架结构分析的自动化有限元建模。 large language model
7 Guided Query Refinement: Multimodal Hybrid Retrieval with Test-Time Optimization 提出引导式查询优化（GQR），提升视觉文档检索效率与性能。 multimodal ✅
8 FedSRD: Sparsify-Reconstruct-Decompose for Communication-Efficient Federated Large Language Models Fine-Tuning 提出FedSRD框架，通过稀疏化-重构-分解，解决联邦LLM微调中的通信瓶颈问题。 large language model
9 Large Language Models Preserve Semantic Isotopies in Story Continuations 研究表明，大型语言模型在故事续写中能够保持语义同位素 large language model
10 SwiReasoning: Switch-Thinking in Latent and Explicit for Pareto-Superior Reasoning LLMs 提出SwiReasoning，通过显隐式推理切换，提升LLM推理性能和效率。 large language model chain-of-thought
11 Mind Your Tone: Investigating How Prompt Politeness Affects LLM Accuracy (short paper) 探究提示语礼貌程度对LLM准确率的影响：不礼貌提示效果更佳 large language model
12 WeatherArchive-Bench: Benchmarking Retrieval-Augmented Reasoning for Historical Weather Archives 提出WeatherArchive-Bench，用于评估历史天气档案的检索增强推理能力。 large language model
13 RAG Makes Guardrails Unsafe? Investigating Robustness of Guardrails under RAG-style Contexts 研究表明RAG上下文增强会降低LLM安全防护模型的可靠性 large language model
14 TeachLM: Post-Training LLMs for Education Using Authentic Learning Data TeachLM：利用真实学习数据后训练LLM，提升教育领域应用效果 large language model
15 Do LLMs Align with My Task? Evaluating Text-to-SQL via Dataset Alignment 通过数据集对齐评估LLM在Text-to-SQL任务中的性能 large language model
16 SimulatorArena: Are User Simulators Reliable Proxies for Multi-Turn Evaluation of AI Assistants? 提出SimulatorArena以评估用户模拟器在AI助手多轮对话中的可靠性 large language model
17 Camellia: Benchmarking Cultural Biases in LLMs for Asian Languages Camellia：用于评估LLM在亚洲语言中文化偏见的基准测试 large language model
18 Proactive defense against LLM Jailbreak ProAct：一种针对LLM越狱攻击的主动防御框架 large language model
19 Retrieval-Augmented Code Generation: A Survey with Focus on Repository-Level Approaches 综述检索增强代码生成，聚焦于仓库级别代码生成方法 large language model
20 When Models Lie, We Learn: Multilingual Span-Level Hallucination Detection with PsiloQA 提出PsiloQA，一个多语言跨度级幻觉检测数据集，并评估多种检测方法。 large language model
21 Instability in Downstream Task Performance During LLM Pretraining 通过检查点集成提升LLM预训练下游任务性能的稳定性 large language model
22 Are BabyLMs Deaf to Gricean Maxims? A Pragmatic Evaluation of Sample-efficient Language Models 提出Gricean Maxims基准，评估小规模语言模型（BabyLMs）的语用推理能力。 large language model
23 A Low-Resource Speech-Driven NLP Pipeline for Sinhala Dyslexia Assistance 提出一套面向僧伽罗语阅读障碍辅助的低资源语音驱动NLP流水线 multimodal
24 A novel hallucination classification framework 提出一种新型幻觉分类框架，用于自动检测大语言模型推理过程中的幻觉。 large language model
25 GenQuest: An LLM-based Text Adventure Game for Language Learners GenQuest：基于LLM的文本冒险游戏，助力语言学习 large language model
26 On the Role of Unobserved Sequences on Sample-based Uncertainty Quantification for LLMs 强调未观测序列在LLM不确定性量化中的作用，并建议未来研究纳入考虑。 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🔗⭐
27 The Geometry of Truth: Layer-wise Semantic Dynamics for Hallucination Detection in Large Language Models 提出Layer-wise Semantic Dynamics (LSD)用于检测大语言模型中的幻觉问题 contrastive learning large language model
28 Mitigating Forgetting Between Supervised and Reinforcement Learning Yields Stronger Reasoners 提出一种高效的SFT与RL集成框架，提升LLM的推理能力并缓解灾难性遗忘。 reinforcement learning large language model chain-of-thought
29 Let it Calm: Exploratory Annealed Decoding for Verifiable Reinforcement Learning 提出探索性退火解码(EAD)，提升可验证强化学习中LLM的推理能力。 reinforcement learning large language model
30 Detecting Distillation Data from Reasoning Models 提出Token Probability Deviation方法，用于检测推理模型蒸馏数据中的污染样本。 distillation large language model
31 Self-Filtered Distillation with LLMs-generated Trust Indicators for Reliable Patent Classification 提出自过滤蒸馏框架，利用LLM生成的可信度指标提升专利分类可靠性 distillation large language model
32 GRACE: Generative Representation Learning via Contrastive Policy Optimization GRACE：通过对比策略优化进行生成式表示学习，提升LLM文本编码能力。 representation learning large language model ✅
33 Aligning Language Models with Clinical Expertise: DPO for Heart Failure Nursing Documentation in Critical Care 利用DPO对Mistral-7B进行微调，提升ICU中护理文档的心力衰竭专业性。 DPO direct preference optimization
34 Hybrid Architectures for Language Models: Systematic Analysis and Design Insights 系统分析混合语言模型架构，优化长文本建模的效率与性能。 Mamba state space model large language model
35 TiTok: Transfer Token-level Knowledge via Contrastive Excess to Transplant LoRA TiTok：利用对比差异迁移token级知识，实现LoRA跨模型移植 distillation large language model
36 Multi-Agent Tool-Integrated Policy Optimization 提出MATPO，通过多智能体强化学习优化工具集成的大语言模型 reinforcement learning large language model
37 Contrastive Learning Using Graph Embeddings for Domain Adaptation of Language Models in the Process Industry 利用图嵌入对比学习，提升语言模型在流程工业领域自适应能力 contrastive learning

🔬 支柱一：机器人控制 (Robot Control) (1 篇)

#题目一句话要点标签🔗⭐
38 JSON Whisperer: Efficient JSON Editing with LLMs JSON Whisperer：利用LLM高效编辑JSON，通过生成diff补丁而非完全重构。 manipulation large language model ✅

⬅️ 返回 cs.CL 首页 · 🏠 返回主页