cs.CL（2025-02-19）

📊 共 67 篇论文 | 🔗 14 篇有代码

🎯 兴趣领域导航

支柱九：具身大模型 (Embodied Foundation Models) (57 🔗12) 支柱二：RL算法与架构 (RL & Architecture) (9 🔗2) 支柱一：机器人控制 (Robot Control) (1)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (57 篇)

#题目一句话要点标签🔗⭐
1 MM-Verify: Enhancing Multimodal Reasoning with Chain-of-Thought Verification 提出MM-Verify，通过思维链验证增强多模态推理能力，超越GPT-4o。 large language model multimodal chain-of-thought
2 ProMedTS: A Self-Supervised, Prompt-Guided Multimodal Approach for Integrating Medical Text and Time Series ProMedTS：一种自监督提示引导的多模态方法，用于整合医学文本和时间序列数据 large language model multimodal
3 Adapting Large Language Models for Time Series Modeling via a Novel Parameter-efficient Adaptation Method Time-LlaMA：一种高效参数适配方法，用于大型语言模型在时间序列建模中的应用 large language model foundation model
4 Latent Distribution Decoupling: A Probabilistic Framework for Uncertainty-Aware Multimodal Emotion Recognition 提出LDDU框架，通过解耦潜在情绪分布建模不确定性，提升多模态情感识别性能。 multimodal ✅
5 Why Safeguarded Ships Run Aground? Aligned Large Language Models' Safety Mechanisms Tend to Be Anchored in The Template Region 揭示大语言模型安全机制脆弱性：模板锚定导致的安全对齐易受攻击 large language model
6 Zero-Shot Commonsense Validation and Reasoning with Large Language Models: An Evaluation on SemEval-2020 Task 4 Dataset 利用大型语言模型进行零样本常识验证与推理：SemEval-2020 Task 4数据集评估 large language model
7 REFIND at SemEval-2025 Task 3: Retrieval-Augmented Factuality Hallucination Detection in Large Language Models 提出REFIND框架，通过检索增强和上下文敏感度量化，检测大语言模型中的幻觉。 large language model ✅
8 Complex Ontology Matching with Large Language Model Embeddings 提出一种融合大型语言模型嵌入的复杂本体匹配方法，显著提升了匹配的表达能力。 large language model
9 What are Models Thinking about? Understanding Large Language Model Hallucinations "Psychology" through Model Inner State Analysis 通过分析LLM内部状态理解幻觉“心理”，实现无需外部信息的幻觉检测。 large language model
10 Batayan: A Filipino NLP benchmark for evaluating Large Language Models Batayan：构建菲律宾语NLP基准，评估大型语言模型在低资源语言上的性能 large language model
11 GIMMICK -- Globally Inclusive Multimodal Multitask Cultural Knowledge Benchmarking GIMMICK：构建全球包容的多模态多任务文化知识基准评测体系 multimodal
12 PRIV-QA: Privacy-Preserving Question Answering for Cloud Large Language Models 提出PRIV-QA：一种面向云端大语言模型的隐私保护问答框架 large language model ✅
13 Detecting Linguistic Bias in Government Documents Using Large language Models 提出DGDB数据集并微调BERT模型，用于检测政府文件中存在的语言偏见。 large language model
14 Towards Lightweight, Adaptive and Attribute-Aware Multi-Aspect Controllable Text Generation with Large Language Models 提出轻量级、自适应和属性感知的多方面可控文本生成框架，提升大语言模型控制能力。 large language model
15 Event Segmentation Applications in Large Language Model Enabled Automated Recall Assessments 利用大语言模型实现自动化事件分割与记忆评估，提升认知研究效率。 large language model
16 MMTEB: Massive Multilingual Text Embedding Benchmark 提出大规模多语言文本嵌入基准MMTEB，用于全面评估文本嵌入模型。 large language model instruction following
17 OpenSearch-SQL: Enhancing Text-to-SQL with Dynamic Few-shot and Consistency Alignment OpenSearch-SQL：通过动态Few-shot和一致性对齐增强Text-to-SQL性能 large language model instruction following
18 STaR-SQL: Self-Taught Reasoner for Text-to-SQL 提出STaR-SQL，通过自学习推理提升Text-to-SQL任务性能 large language model chain-of-thought
19 Transferring Textual Preferences to Vision-Language Understanding through Model Merging 通过模型融合将文本偏好迁移到视觉-语言理解模型 multimodal
20 RocketKV: Accelerating Long-Context LLM Inference via Two-Stage KV Cache Compression RocketKV：通过两阶段KV缓存压缩加速长文本LLM推理。 large language model ✅
21 DataSciBench: An LLM Agent Benchmark for Data Science DataSciBench：一个用于评估LLM在数据科学任务中能力的综合性基准测试。 large language model ✅
22 The Canary's Echo: Auditing Privacy Risks of LLM-Generated Synthetic Text 针对LLM生成合成文本的隐私风险，提出新型数据驱动的成员推理攻击方法。 large language model
23 VITAL: A New Dataset for Benchmarking Pluralistic Alignment in Healthcare VITAL：针对医疗领域多元化对齐的基准数据集 large language model
24 MaskPrune: Mask-based LLM Pruning for Layer-wise Uniform Structures 提出MaskPrune，通过掩码学习实现LLM逐层均匀结构化剪枝 large language model
25 PLDR-LLMs Learn A Generalizable Tensor Operator That Can Replace Its Own Deep Neural Net At Inference 提出PLDR-LLM以替代深度神经网络进行推理 large language model
26 ThinkGuard: Deliberative Slow Thinking Leads to Cautious Guardrails ThinkGuard：通过审慎的慢思考实现更可靠的大语言模型安全防护 large language model
27 A Tale of Two Structures: Do LLMs Capture the Fractal Complexity of Language? 研究表明大型语言模型在多大程度上能捕捉语言的分形复杂性，并揭示了影响因素。 large language model
28 How Do LLMs Perform Two-Hop Reasoning in Context? 揭示LLM上下文学习中双跳推理的机制：从随机猜测到结构化查询 large language model
29 SIFT: Grounding LLM Reasoning in Contexts via Stickers SIFT：通过Sticker机制增强LLM在上下文中的推理能力 large language model ✅
30 C2T: A Classifier-Based Tree Construction Method in Speculative Decoding 提出C2T方法，利用分类器动态构建token树，提升推理解码效率。 large language model
31 Activation-aware Probe-Query: Effective Key-Value Retrieval for Long-Context LLMs Inference 提出ActQKV以解决长上下文LLMs推理效率问题 large language model
32 Towards Geo-Culturally Grounded LLM Generations 检索增强提升LLM文化感知能力，但需警惕刻板印象风险 large language model
33 EvoP: Robust LLM Inference via Evolutionary Pruning EvoP：通过演化剪枝实现鲁棒的大语言模型推理 large language model
34 TreeCut: A Synthetic Unanswerable Math Word Problem Dataset for LLM Hallucination Evaluation TreeCut：用于评估LLM幻觉的合成不可解数学应用题数据集 large language model ✅
35 The Self-Improvement Paradox: Can Language Models Bootstrap Reasoning Capabilities without External Scaffolding? Crescent：一种无需外部监督信号的LLM自提升数学推理能力框架 large language model
36 Benchmarking LLMs for Political Science: A United Nations Perspective 提出联合国基准以评估大语言模型在政治决策中的应用 large language model ✅
37 Towards Context-Robust LLMs: A Gated Representation Fine-tuning Approach 提出Grft门控微调方法，提升LLM在检索增强生成中对上下文的鲁棒性 large language model
38 Retrieving Versus Understanding Extractive Evidence in Few-Shot Learning 分析少样本学习中LLM抽取式证据检索与理解的关联性，揭示模型预测误差与证据检索误差的关系。 large language model
39 Are Rules Meant to be Broken? Understanding Multilingual Moral Reasoning as a Computational Pipeline with UniMoral UniMoral：构建多语言道德推理计算流程，弥合文化差异下的道德理解鸿沟 large language model
40 Beyond Single Frames: Can LMMs Comprehend Temporal and Contextual Narratives in Image Sequences? 提出StripCipher基准，评估LMMs在图像序列中的时序和上下文理解能力 multimodal
41 TESS 2: A Large-Scale Generalist Diffusion Language Model TESS 2：一种大规模通用扩散语言模型，性能媲美自回归模型。 instruction following ✅
42 Inner Thinking Transformer: Leveraging Dynamic Depth Scaling to Foster Adaptive Internal Thinking 提出Inner Thinking Transformer，通过动态深度缩放提升LLM在关键token上的推理能力。 large language model
43 From Tools to Teammates: Evaluating LLMs in Multi-Session Coding Interactions MemoryCode：评估LLM在多轮编码交互中的长期记忆能力 large language model
44 SCALAR: Scientific Citation-based Live Assessment of Long-context Academic Reasoning 提出SCALAR：基于科学引用的长文本学术推理实时评估基准 large language model
45 Enhancing Input-Label Mapping in In-Context Learning with Contrastive Decoding 提出In-Context Contrastive Decoding，增强ICL中输入-标签映射，提升NLU任务性能。 large language model ✅
46 Is This Collection Worth My LLM's Time? Automatically Measuring Information Potential in Text Corpora 提出一种无需训练的文本信息价值评估方法，助力大语言模型高效数据集成 large language model
47 Instruction Tuning on Public Government and Cultural Data for Low-Resource Language: a Case Study in Kazakh 针对低资源语言哈萨克语，提出基于政府和文化数据的指令调优方法。 instruction following
48 D.Va: Validate Your Demonstration First Before You Use It 提出D.Va：一种基于验证的ICL示例选择方法，提升LLM在NLU/NLG任务上的性能。 large language model
49 Qorgau: Evaluating LLM Safety in Kazakh-Russian Bilingual Contexts Qorgau：评估LLM在哈萨克-俄语双语环境下的安全性 large language model
50 BeamLoRA: Beam-Constraint Low-Rank Adaptation BeamLoRA：提出一种基于Beam搜索的低秩自适应方法，提升大语言模型微调精度。 large language model
51 Shall Your Data Strategy Work? Perform a Swift Study 提出一种快速评估指令微调数据有效性的方法，无需模型重训练。 chain-of-thought
52 Detecting LLM Fact-conflicting Hallucinations Enhanced by Temporal-logic-based Reasoning Drowzee：利用时序逻辑增强LLM事实冲突幻觉检测 large language model
53 Prompting a Weighting Mechanism into LLM-as-a-Judge in Two-Step: A Case Study 提出一种基于Prompt的权重机制，提升LLM作为评判者的NLG任务评估能力 large language model
54 Reducing Hallucinations in Language Model-based SPARQL Query Generation Using Post-Generation Memory Retrieval 提出PGMR框架，利用后生成记忆检索减少LLM在SPARQL查询生成中的幻觉问题 large language model
55 Bridging the Editing Gap in LLMs: FineEdit for Precise and Targeted Text Modifications FineEdit：通过微调LLM实现精准和目标明确的文本编辑 large language model ✅
56 Craw4LLM: Efficient Web Crawling for LLM Pretraining Craw4LLM：面向LLM预训练的高效网页爬取方法，显著降低爬取浪费。 large language model ✅
57 GneissWeb: Preparing High Quality Data for LLMs at Scale GneissWeb：构建高质量大规模LLM训练数据集，提升模型泛化能力 large language model

🔬 支柱二：RL算法与架构 (RL & Architecture) (9 篇)

#题目一句话要点标签🔗⭐
58 Direct Value Optimization: Improving Chain-of-Thought Reasoning in LLMs with Refined Values 提出直接价值优化(DVO)，通过精细化价值信号提升LLM的CoT推理能力 reinforcement learning large language model chain-of-thought
59 LongPO: Long Context Self-Evolution of Large Language Models through Short-to-Long Preference Optimization LongPO：通过短到长偏好优化实现大语言模型长上下文自进化 DPO large language model ✅
60 Refining Sentence Embedding Model through Ranking Sentences Generation with Large Language Models 提出基于大语言模型排序生成句子的句子嵌入模型优化方法 contrastive learning large language model
61 LLM should think and action as a human 提出基于内置思维链的LLM交互方法，提升多轮对话中的推理和规划能力 reinforcement learning large language model chain-of-thought
62 MuDAF: Long-Context Multi-Document Attention Focusing through Contrastive Learning on Attention Heads 提出MuDAF，通过对比学习优化注意力头，提升长文本多文档问答性能。 contrastive learning large language model
63 RLTHF: Targeted Human Feedback for LLM Alignment RLTHF：通过有针对性的人工反馈实现LLM对齐，降低标注成本并提升性能。 reinforcement learning RLHF large language model
64 Task-agnostic Prompt Compression with Context-aware Sentence Embedding and Reward-guided Task Descriptor 提出任务无关提示压缩框架TPC，提升LLM在长文本任务中的泛化能力。 reinforcement learning large language model
65 MoM: Linear Sequence Modeling with Mixture-of-Memories 提出MoM：一种混合记忆的线性序列建模方法，提升长序列召回能力 state space model linear attention ✅
66 Efficient Safety Retrofitting Against Jailbreaking for LLMs 提出Egida数据集和DPO微调方法，高效提升LLM抗越狱攻击的安全性 DPO direct preference optimization

🔬 支柱一：机器人控制 (Robot Control) (1 篇)

#题目一句话要点标签🔗⭐
67 Don't Stop the Multi-Party! On Generating Synthetic Multi-Party Conversations with Constraints 提出基于指令微调LLM的受约束多方对话生成方法，解决现有数据集隐私和平台局限性问题。 MPC large language model

⬅️ 返回 cs.CL 首页 · 🏠 返回主页