cs.CL（2026-02-06）

📊 共 27 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) (16 🔗2) 支柱二：RL算法与架构 (RL & Architecture) (10 🔗3) 支柱六：视频提取与匹配 (Video Extraction) (1)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (16 篇)

#题目一句话要点标签🔗⭐
1 Beyond Static Alignment: Hierarchical Policy Control for LLM Safety via Risk-Aware Chain-of-Thought PACT：通过风险感知的思维链实现LLM安全性的分层策略控制 large language model chain-of-thought
2 CORE: Comprehensive Ontological Relation Evaluation for Large Language Models CORE：用于评估大型语言模型本体关系理解能力的综合数据集 large language model
3 Improve Large Language Model Systems with User Logs 提出UNO框架，利用用户日志提升大语言模型系统性能 large language model ✅
4 Comprehensive Evaluation of Large Language Models on Software Engineering Tasks: A Multi-Task Benchmark 多任务基准测试全面评估大语言模型在软件工程任务中的表现 large language model
5 Not All Layers Need Tuning: Selective Layer Restoration Recovers Diversity 提出选择性层恢复(SLR)方法，解决LLM后训练中生成多样性降低的问题。 large language model instruction following
6 Visual Word Sense Disambiguation with CLIP through Dual-Channel Text Prompting and Image Augmentations 提出基于CLIP和双通道文本提示的视觉词义消歧框架，提升词义理解能力。 large language model multimodal
7 Evaluating Prompt Engineering Strategies for Sentiment Control in AI-Generated Texts 提出基于Prompt工程的情感控制方法，提升AI生成文本的情感表达能力 large language model chain-of-thought
8 Uncovering Cross-Objective Interference in Multi-Objective Alignment 揭示多目标对齐中跨目标干扰现象，提出CTWA方法缓解性能退化。 large language model
9 Quantum Attention by Overlap Interference: Predicting Sequences from Classical and Many-Body Quantum Data 提出基于重叠干涉的量子注意力机制，用于预测经典和量子多体数据序列。 large language model
10 Inference-Time Rethinking with Latent Thought Vectors for Math Reasoning 提出基于隐变量思维向量的推理时反思框架，提升数学推理能力。 chain-of-thought
11 Your Language Model Secretly Contains Personality Subnetworks 揭示大语言模型中隐藏的个性子网络，实现无需训练的个性化控制。 large language model
12 Massive Sound Embedding Benchmark (MSEB) 提出大规模声音嵌入基准MSEB，用于评估多模态系统中音频理解能力 multimodal
13 DAWN: Dependency-Aware Fast Inference for Diffusion LLMs DAWN：面向Diffusion LLM的依赖感知快速推理方法，提升解码效率。 large language model ✅
14 Do Prompts Guarantee Safety? Mitigating Toxicity from LLM Generations through Subspace Intervention 提出基于子空间干预的LLM解毒方法，有效降低生成文本中的毒性。 large language model
15 Cost-Aware Model Selection for Text Classification: Multi-Objective Trade-offs Between Fine-Tuned Encoders and LLM Prompting in Production 针对文本分类，提出一种考虑成本的多目标模型选择方法，平衡微调编码器与LLM Prompting。 large language model
16 Lost in Speech: Benchmarking, Evaluation, and Parsing of Spoken Code-Switching Beyond Standard UD Assumptions 针对口语代码切换，提出DECAP框架和FLEX-UD评估指标，提升句法分析性能。 large language model

🔬 支柱二：RL算法与架构 (RL & Architecture) (10 篇)

#题目一句话要点标签🔗⭐
17 Evaluating an evidence-guided reinforcement learning framework in aligning light-parameter large language models with decision-making cognition in psychiatric clinical reasoning ClinMPO：证据引导强化学习提升轻量级LLM在精神病学临床推理中的决策认知能力 reinforcement learning large language model
18 FMBench: Adaptive Large Language Model Output Formatting FMBench：自适应大语言模型Markdown格式化输出评测与优化 reinforcement learning large language model instruction following ✅
19 compar:IA: The French Government's LLM arena to collect French-language human prompts and preference data compar:IA：法国政府构建法语LLM评测平台，收集人类prompt和偏好数据 reinforcement learning RLHF DPO
20 R-Align: Enhancing Generative Reward Models through Rationale-Centric Meta-Judging R-Align：通过以推理为中心的元判断增强生成式奖励模型 reinforcement learning RLHF large language model
21 InftyThink+: Effective and Efficient Infinite-Horizon Reasoning via Reinforcement Learning InftyThink+：通过强化学习实现高效无限视野推理 reinforcement learning chain-of-thought
22 TrailBlazer: History-Guided Reinforcement Learning for Black-Box LLM Jailbreaking 提出历史引导的强化学习框架以提升黑箱LLM越狱效率 reinforcement learning large language model
23 SEMA: Simple yet Effective Learning for Multi-Turn Jailbreak Attacks 提出SEMA框架，通过自调优预填充和意图感知强化学习，有效提升多轮对抗攻击成功率。 reinforcement learning DPO direct preference optimization ✅
24 Can Post-Training Transform LLMs into Causal Reasoners? 通过后训练将大语言模型转化为因果推理器 PPO DPO large language model ✅
25 Generating Data-Driven Reasoning Rubrics for Domain-Adaptive Reward Modeling 提出数据驱动的推理评估准则，提升领域自适应奖励建模效果 reinforcement learning large language model
26 Free Energy Mixer 提出自由能混合器(FEM)，通过值驱动的通道选择提升注意力机制性能。 SSM linear attention

🔬 支柱六：视频提取与匹配 (Video Extraction) (1 篇)

#题目一句话要点标签🔗⭐
27 On the Wings of Imagination: Conflicting Script-based Multi-role Framework for Humor Caption Generation 提出基于冲突脚本的多角色框架HOMER，用于生成幽默的图像描述 HuMoR large language model

⬅️ 返回 cs.CL 首页 · 🏠 返回主页