cs.CL（2025-03-03）

📊 共 26 篇论文 | 🔗 3 篇有代码

🎯 兴趣领域导航

支柱九：具身大模型 (Embodied Foundation Models) (22 🔗2) 支柱四：生成式动作 (Generative Motion) (1) 支柱二：RL算法与架构 (RL & Architecture) (1) 支柱七：动作重定向 (Motion Retargeting) (1 🔗1) 支柱一：机器人控制 (Robot Control) (1)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (22 篇)

#题目一句话要点标签🔗⭐
1 HoT: Highlighted Chain of Thought for Referencing Supporting Facts from Inputs 提出高亮思维链（HoT）提示方法，提升LLM事实依据追溯能力并辅助人工验证。 large language model chain-of-thought
2 Linear Representations of Political Perspective Emerge in Large Language Models 大型语言模型中涌现政治立场的线性表征，可通过干预注意力头操控模型输出。 large language model
3 Analyzing the Safety of Japanese Large Language Models in Stereotype-Triggering Prompts 分析日语大型语言模型在刻板印象触发提示下的安全性 large language model
4 EAGLE-3: Scaling up Inference Acceleration of Large Language Models via Training-Time Test EAGLE-3：通过训练时测试扩展大语言模型推理加速，提升数据规模利用率 large language model ✅
5 Rotary Offset Features in Large Language Models 揭示LLM中Rotary Embedding的Offset Features，并提供预测方法 large language model
6 Persuade Me if You Can: A Framework for Evaluating Persuasion Effectiveness and Susceptibility Among Large Language Models PMIYC：评估大型语言模型说服力及易受说服性的自动化框架 large language model
7 Retrieval Models Aren't Tool-Savvy: Benchmarking Tool Retrieval for Large Language Models 提出ToolRet基准评测工具检索模型，并构建大规模训练数据集提升LLM工具使用能力。 large language model
8 Phi-4-Mini Technical Report: Compact yet Powerful Multimodal Language Models via Mixture-of-LoRAs 微软发布Phi-4-Mini系列模型，通过混合LoRA实现紧凑而强大的多模态语言能力。 multimodal
9 Building Safe GenAI Applications: An End-to-End Overview of Red Teaming for Large Language Models 针对大型语言模型的红队评估综述：构建安全的GenAI应用 large language model
10 Automated Annotation of Evolving Corpora for Augmenting Longitudinal Network Data: A Framework Integrating Large Language Models and Expert Knowledge 提出EALA框架，结合LLM与专家知识，自动标注演化语料以增强纵向网络数据。 large language model
11 Detecting Stylistic Fingerprints of Large Language Models 提出一种基于集成学习的LLM风格指纹检测方法，用于识别AI生成文本的来源。 large language model
12 What do Large Language Models Say About Animals? Investigating Risks of Animal Harm in Generated Text 提出AnimalHarmBench基准，评估大型语言模型生成文本中潜在的动物伤害风险。 large language model
13 CrowdSelect: Synthetic Instruction Data Selection with Multi-LLM Wisdom CrowdSelect：利用多LLM智慧进行合成指令数据选择，提升小模型指令遵循能力。 large language model instruction following ✅
14 Interview AI-ssistant: Designing for Real-Time Human-AI Collaboration in Interview Preparation and Execution 提出 Interview AI-ssistant，用于访谈准备和执行中的人机实时协作 large language model
15 Persuasion at Play: Understanding Misinformation Dynamics in Demographic-Aware Human-LLM Interactions 研究人口统计学背景下人-LLM交互中的错误信息传播动态 large language model
16 Comparative Analysis of OpenAI GPT-4o and DeepSeek R1 for Scientific Text Categorization Using Prompt Engineering 利用提示工程比较OpenAI GPT-4o和DeepSeek R1在科学文本分类中的性能 large language model
17 Mind the (Belief) Gap: Group Identity in the World of LLMs 研究LLM中的群体认同偏差，提出干预策略以减少信息误传并提升学习效果 large language model
18 Can (A)I Change Your Mind? 研究表明：大型语言模型在希伯来语环境下能有效改变人类观点 large language model
19 From Language to Cognition: How LLMs Outgrow the Human Language Network 研究表明LLM的语言能力发展与人脑语言网络关联，但超越人类后关联减弱 large language model
20 $\texttt{SEM-CTRL}$: Semantically Controlled Decoding 提出SEM-CTRL，通过语义控制解码保证LLM输出的句法和语义正确性 large language model
21 Word Form Matters: LLMs' Semantic Reconstruction under Typoglycemia 研究发现LLM在乱序词理解中过度依赖词形，并提出SemRecScore评估语义重构能力 large language model
22 Evaluating LLMs' Assessment of Mixed-Context Hallucination Through the Lens of Summarization 通过摘要任务评估LLMs对混合上下文幻觉的检测能力 large language model

🔬 支柱四：生成式动作 (Generative Motion) (1 篇)

#题目一句话要点标签🔗⭐
23 Superscopes: Amplifying Internal Feature Representations for Language Model Interpretation Superscopes：通过放大内部特征表示增强语言模型的可解释性 classifier-free guidance large language model

🔬 支柱二：RL算法与架构 (RL & Architecture) (1 篇)

#题目一句话要点标签🔗⭐
24 The Emergence of Grammar through Reinforcement Learning 提出基于强化学习的语法演化模型，探索语言表达目的对语法形成的影响 reinforcement learning

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

#题目一句话要点标签🔗⭐
25 Why Is Spatial Reasoning Hard for VLMs? An Attention Mechanism Perspective on Focus Areas 提出ADAPTVIS，通过置信度调整视觉注意力，显著提升VLM空间推理能力 spatial relationship ✅

🔬 支柱一：机器人控制 (Robot Control) (1 篇)

#题目一句话要点标签🔗⭐
26 Twenty Years of Personality Computing: Threats, Challenges and Future Directions 人格计算二十年：回顾、挑战与未来发展方向 manipulation

⬅️ 返回 cs.CL 首页 · 🏠 返回主页