cs.CL（2025-09-27）

📊 共 34 篇论文 | 🔗 5 篇有代码

🎯 兴趣领域导航

支柱九：具身大模型 (Embodied Foundation Models) (27 🔗4) 支柱二：RL算法与架构 (RL & Architecture) (7 🔗1)

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

#题目一句话要点标签🔗⭐
1 Comparison of Scoring Rationales Between Large Language Models and Human Raters 对比大型语言模型与人类评分者的评分理由，探究自动评分一致性问题 large language model
2 How to Make Large Language Models Generate 100% Valid Molecules? 提出SmiSelf框架，确保大语言模型生成100%有效分子 large language model ✅
3 The Impact of Role Design in In-Context Learning for Large Language Models 研究角色设计对大语言模型上下文学习的影响，提升零样本和少样本学习性能 large language model
4 Cognition-of-Thought Elicits Social-Aligned Reasoning in Large Language Models 提出CooT：一种基于认知自监控的大语言模型社会对齐推理框架 large language model
5 Detecting Corpus-Level Knowledge Inconsistencies in Wikipedia with Large Language Models 提出CLAIRE系统，检测维基百科语料库级别知识不一致性，提升编辑效率。 large language model
6 A Structured Framework for Evaluating and Enhancing Interpretive Capabilities of Multimodal LLMs in Culturally Situated Tasks 构建结构化框架，提升多模态LLM在中国文化情境下的理解能力 multimodal ✅
7 CCD: Mitigating Hallucinations in Radiology MLLMs via Clinical Contrastive Decoding 提出临床对比解码以解决放射学多模态大语言模型的幻觉问题 large language model multimodal
8 Modeling the language cortex with form-independent and enriched representations of sentence meaning reveals remarkable semantic abstractness 提出基于视觉和语言模型的抽象语义表示以揭示语言皮层特征 large language model
9 An Senegalese Legal Texts Structuration Using LLM-augmented Knowledge Graph 利用LLM增强的知识图谱构建塞内加尔法律文本结构化方法，提升法律信息可访问性。 large language model
10 Train Once, Answer All: Many Pretraining Experiments for the Cost of One 提出单次训练多重实验方法，降低大语言模型预训练实验成本。 large language model
11 Alignment through Meta-Weighted Online Sampling: Bridging the Gap between Data Generation and Preference Optimization 提出MetaAPO，通过元加权在线采样弥合数据生成与偏好优化间的差距 large language model
12 Breaking the MoE LLM Trilemma: Dynamic Expert Clustering with Structured Compression 提出基于动态专家聚类与结构化压缩的MoE LLM优化框架，解决负载不均衡、参数冗余和通信开销问题。 large language model
13 Evaluating Bias in Spoken Dialogue LLMs for Real-World Decisions and Recommendations 系统评估语音对话大模型在决策和推荐中的偏见，揭示多轮对话的影响。 large language model
14 Language, Culture, and Ideology: Personalizing Offensiveness Detection in Political Tweets with Reasoning LLMs 利用推理LLM个性化政治推文冒犯性检测，考虑语言、文化和意识形态因素 large language model
15 Dual-Space Smoothness for Robust and Balanced LLM Unlearning PRISM：通过双空间平滑实现鲁棒且均衡的大语言模型不可学习 large language model
16 LLMSQL: Upgrading WikiSQL for the LLM Era of Text-to-SQL LLMSQL：为大语言模型时代升级WikiSQL文本到SQL数据集 large language model
17 A2D: Any-Order, Any-Step Safety Alignment for Diffusion Language Models A2D：针对扩散语言模型的任意顺序、任意步骤安全对齐方法 large language model
18 Small Language Models for Curriculum-based Guidance 利用小型语言模型和课程指导，构建可持续的AI教学助手 large language model
19 $\texttt{BluePrint}$: A Social Media User Dataset for LLM Persona Evaluation and Training 提出BluePrint数据集，用于评估和训练LLM在社交媒体用户建模中的Persona表现。 large language model
20 MaskSQL: Safeguarding Privacy for LLM-Based Text-to-SQL via Abstraction 提出MaskSQL以解决LLM文本到SQL的隐私保护问题 large language model
21 No Loss, No Gain: Gated Refinement and Adaptive Compression for Prompt Optimization 提出GRACE框架，通过门控优化和自适应压缩提升Prompt优化效率与性能。 large language model ✅
22 PonderLM-2: Pretraining LLM with Latent Thoughts in Continuous Space PonderLM-2：通过在连续空间中预训练具有潜在思想的LLM，提升单token生成质量。 chain-of-thought
23 Test-Time Policy Adaptation for Enhanced Multi-Turn Interactions with LLMs 提出测试时策略自适应(T2PAM)框架，增强LLM多轮交互中的性能。 large language model
24 d$^2$Cache: Accelerating Diffusion-Based LLMs via Dual Adaptive Caching 提出d$^2$Cache，通过双重自适应缓存加速基于扩散的LLM推理。 large language model ✅
25 The Geometry of Creative Variability: How Credal Sets Expose Calibration Gaps in Language Models 利用Credal集几何分析揭示语言模型在创意生成中校准差距 large language model
26 Semantic Voting: A Self-Evaluation-Free Approach for Efficient LLM Self-Improvement on Unverifiable Open-ended Tasks 提出语义投票方法，无需自评估即可高效提升LLM在不可验证开放任务上的性能。 large language model
27 Peacemaker or Troublemaker: How Sycophancy Shapes Multi-Agent Debate 提出多智能体辩论中谄媚行为的评估框架，揭示其对辩论质量的负面影响 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 DRIFT: Learning from Abundant User Dissatisfaction in Real-World Preference Learning DRIFT：利用真实用户不满意信号进行偏好学习，提升大语言模型性能 preference learning DPO large language model ✅
29 MedCritical: Enhancing Medical Reasoning in Small Language Models via Self-Collaborative Correction MedCritical：通过自协同校正增强小语言模型在医疗推理中的能力 DPO direct preference optimization distillation
30 Learning to Reason in Structured In-context Environments with Reinforcement Learning 提出SIE框架，利用强化学习提升LLM在结构化上下文环境中的推理能力 reinforcement learning large language model
31 Tagging the Thought: Unlocking Personalization Reasoning via Reinforcement Learning TagPR：通过强化学习和思维标注提升LLM的个性化推理能力 reinforcement learning large language model
32 From Evidence to Trajectory: Abductive Reasoning Path Synthesis for Training Retrieval-Augmented Generation Agents 提出EviPath，通过证据推理路径合成训练RAG Agent，提升开放域问答性能。 reinforcement learning large language model chain-of-thought
33 Look Back to Reason Forward: Revisitable Memory for Long-Context LLM Agents 提出ReMemR1，通过可回顾记忆增强长文本LLM Agent的推理能力 reinforcement learning reward design large language model
34 PARL-MT: Learning to Call Functions in Multi-Turn Conversation with Progress Awareness PARL-MT：通过进度感知学习在多轮对话中进行函数调用 reinforcement learning large language model

⬅️ 返回 cs.CL 首页 · 🏠 返回主页