cs.CL（2025-02-21）

📊 共 50 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) (44 🔗11) 支柱二：RL算法与架构 (RL & Architecture) (5 🔗3) 支柱一：机器人控制 (Robot Control) (1)

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

#题目一句话要点标签🔗⭐
1 Modality-Aware Neuron Pruning for Unlearning in Multimodal Large Language Models 提出MANU框架，解决多模态大语言模型中的模态感知式遗忘难题 large language model multimodal
2 Chain-of-Rank: Enhancing Large Language Models for Domain-Specific RAG in Edge Device 提出Chain-of-Rank，增强边缘设备上领域特定RAG的大语言模型性能 large language model chain-of-thought
3 Evaluating Multimodal Generative AI with Korean Educational Standards 提出KoNET基准，利用韩国国家教育考试评估多模态生成AI系统 multimodal ✅
4 CoT-ICL Lab: A Synthetic Framework for Studying Chain-of-Thought Learning from In-Context Demonstrations CoT-ICL Lab：用于研究思维链上下文学习的合成框架 chain-of-thought
5 MMRAG: Multi-Mode Retrieval-Augmented Generation with Large Language Models for Biomedical In-Context Learning MMRAG：多模式检索增强生成框架，提升生物医学领域上下文学习效果 large language model
6 ExpliCa: Evaluating Explicit Causal Reasoning in Large Language Models 提出ExpliCa数据集以评估大语言模型的显式因果推理能力 large language model
7 When Compression Meets Model Compression: Memory-Efficient Double Compression for Large Language Models 提出面向量化大语言模型的双重压缩框架，实现内存高效部署 large language model
8 Constructing a Norm for Children's Scientific Drawing: Distribution Features Based on Semantic Similarity of Large Language Models 构建儿童科学绘画规范：基于大语言模型语义相似度的分布特征 large language model
9 Understand User Opinions of Large Language Models via LLM-Powered In-the-Moment User Experience Interviews CLUE：利用LLM进行用户体验访谈，深入理解用户对大型语言模型的观点 large language model ✅
10 Control Illusion: The Failure of Instruction Hierarchies in Large Language Models 揭示大语言模型指令层级控制失效：系统指令易被用户指令覆盖 large language model
11 KVLink: Accelerating Large Language Models via Efficient KV Cache Reuse KVLink：通过高效的KV缓存复用加速大型语言模型 large language model
12 AutoMedPrompt: A New Framework for Optimizing LLM Medical Prompts Using Textual Gradients AutoMedPrompt：利用文本梯度优化LLM医学提示，显著提升医学问答性能 large language model foundation model chain-of-thought
13 CVE-LLM : Ontology-Assisted Automatic Vulnerability Evaluation Using Large Language Models 提出CVE-LLM，利用LLM和本体知识自动评估医疗设备漏洞 large language model
14 Improving Consistency in Large Language Models through Chain of Guidance 提出链式引导（CoG）方法，提升大语言模型在问答任务中的语义一致性。 large language model
15 SafeInt: Shielding Large Language Models from Jailbreak Attacks via Safety-Aware Representation Intervention SafeInt：通过安全感知表示干预防御大语言模型的越狱攻击 large language model
16 SOTOPIA-$Ω$: Dynamic Strategy Injection Learning and Social Instruction Following Evaluation for Social Agents SOTOPIA-$Ω$：动态策略注入学习与社交指令跟随评估，提升社交智能体能力 instruction following
17 Scale-Distribution Decoupling: Enabling Stable and Effective Training of Large Language Models 提出尺度-分布解耦方法，稳定高效地训练大型语言模型 large language model ✅
18 Unveiling Attractor Cycles in Large Language Models: A Dynamical Systems View of Successive Paraphrasing 揭示大语言模型中的吸引子环：基于动态系统视角的连续释义研究 large language model
19 Does Reasoning Introduce Bias? A Study of Social Bias Evaluation and Mitigation in LLM Reasoning 提出ADBP方法，用于缓解LLM推理中由社会偏见导致的错误结论。 large language model chain-of-thought ✅
20 Almost AI, Almost Human: The Challenge of Detecting AI-Polished Writing 评估AI润色文本检测的挑战：现有检测器易误判，且对模型存在偏见 large language model
21 Machine-generated text detection prevents language model collapse 提出基于机器生成文本检测的重要性重采样方法，以防止语言模型崩塌。 large language model
22 Probe Pruning: Accelerating LLMs through Dynamic Pruning via Model-Probing Probe Pruning：通过模型探测实现LLM的动态剪枝加速 large language model ✅
23 On the Robustness of Transformers against Context Hijacking for Linear Classification 研究Transformer在上下文劫持下的鲁棒性，揭示深度模型优势 large language model
24 Do Multilingual LLMs Think In English? 揭示多语言LLM内部决策偏向：模型在英语表征空间进行关键推理 large language model
25 LightThinker: Thinking Step-by-Step Compression LightThinker：提出一种逐步压缩中间推理过程的LLM优化方法。 large language model ✅
26 Synthetic vs. Gold: The Role of LLM Generated Labels and Data in Cyberbullying Detection 利用LLM生成数据和标签，提升网络欺凌检测性能。 large language model
27 AttentionEngine: A Versatile Framework for Efficient Attention Mechanisms on Diverse Hardware Platforms AttentionEngine：用于异构硬件平台高效注意力机制的通用框架 large language model ✅
28 Sparsity May Be All You Need: Sparse Random Parameter Adaptation 提出稀疏随机参数微调方法，在参数效率微调中与LoRA具有竞争力 large language model
29 Generalizing From Short to Long: Effective Data Synthesis for Long-Context Instruction Tuning 提出上下文合成方法，提升短文本指令调优模型在长文本上的泛化能力 large language model ✅
30 ParamMute: Suppressing Knowledge-Critical FFNs for Faithful Retrieval-Augmented Generation ParamMute：抑制知识关键FFN，提升检索增强生成的事实性 large language model ✅
31 Beyond Translation: LLM-Based Data Generation for Multilingual Fact-Checking 提出MultiSynFact：基于LLM的多语言事实核查数据集，支持低资源语言。 large language model
32 Round Attention: A Novel Round-Level Attention Mechanism to Accelerate LLM Inference 提出Round Attention，加速LLM推理并降低KV缓存内存占用。 large language model
33 TETRIS: Optimal Draft Token Selection for Batch Speculative Decoding TETRIS：面向批量推测解码的最优草稿令牌选择方法 large language model
34 Extreme Speech Classification in the Era of LLMs: Exploring Open-Source and Proprietary Models 利用LLM进行极端言论分类：探索开源与闭源模型的性能差异与微调策略 large language model
35 LLMs in Mobile Apps: Practices, Challenges, and Opportunities 构建LLM移动应用数据集，揭示集成策略、挑战与机遇 large language model
36 A Close Look at Decomposition-based XAI-Methods for Transformer Language Models 对比分解式XAI方法，提升Transformer语言模型可解释性 large language model
37 DReSD: Dense Retrieval for Speculative Decoding DReSD：用于推测解码的稠密检索，显著提升LLM生成速度与质量 large language model
38 Stepwise Informativeness Search for Efficient and Effective LLM Reasoning 提出Stepwise Informativeness Search，提升LLM多步推理的准确性和效率 large language model
39 Detecting Future-related Contexts of Entity Mentions 提出实体未来语境检测方法，解决信息处理中自动时序分析的需求。 large language model
40 A Training-free LLM-based Approach to General Chinese Character Error Correction 提出基于LLM的免训练通用中文错别字纠正方法，无需微调即可媲美大模型。 large language model
41 A General Pseudonymization Framework for Cloud-Based LLMs: Replacing Privacy Information in Controlled Text Generation 提出通用伪匿名化框架以解决云端LLMs隐私问题 large language model ✅
42 When Disagreements Elicit Robustness: Investigating Self-Repair Capabilities under LLM Multi-Agent Disagreements 研究LLM多智能体在分歧下的自修复能力，提升复杂任务的鲁棒性 large language model ✅
43 Latent Factor Models Meets Instructions: Goal-conditioned Latent Factor Discovery without Task Supervision Instruct-LF：结合指令LLM与隐因子模型，实现无任务监督的隐概念发现 instruction following
44 Do LLMs Make Mistakes Like Students? Exploring Natural Alignment between Language Models and Human Error Patterns 研究LLM在多选题中犯错模式与学生错误的自然对齐性，为教育评估提供新思路。 large language model

🔬 支柱二：RL算法与架构 (RL & Architecture) (5 篇)

#题目一句话要点标签🔗⭐
45 PPC-GPT: Federated Task-Specific Compression of Large Language Models via Pruning and Chain-of-Thought Distillation PPC-GPT：通过剪枝和CoT蒸馏实现联邦环境下大语言模型的任务特定压缩 distillation large language model chain-of-thought ✅
46 LightMamba: Efficient Mamba Acceleration on FPGA with Quantization and Hardware Co-design LightMamba：通过量化和硬件协同设计，在FPGA上高效加速Mamba模型推理。 Mamba SSM state space model ✅
47 LaTIM: Measuring Latent Token-to-Token Interactions in Mamba Models 提出LaTIM，用于度量Mamba模型中潜在的token间交互，提升模型可解释性。 Mamba SSM state space model
48 Self-Taught Agentic Long Context Understanding 提出AgenticLU框架，通过自学习Agent提升LLM在长文本理解中的问答能力。 direct preference optimization large language model
49 Problem-Solving Logic Guided Curriculum In-Context Learning for LLMs Complex Reasoning 提出问题解决逻辑引导的课程ICL方法，提升LLM的复杂推理能力 curriculum learning large language model ✅

🔬 支柱一：机器人控制 (Robot Control) (1 篇)

#题目一句话要点标签🔗⭐
50 Steering into New Embedding Spaces: Analyzing Cross-Lingual Alignment Induced by Model Interventions in Multilingual Language Models 通过模型干预增强多语言模型跨语言对齐，提升检索性能。 manipulation large language model

⬅️ 返回 cs.CL 首页 · 🏠 返回主页