cs.CL（2026-02-04）

📊 共 34 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) (24 🔗2) 支柱二：RL算法与架构 (RL & Architecture) (10 🔗1)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (24 篇)

#题目一句话要点标签🔗⭐
1 Model-Dowser: Data-Free Importance Probing to Mitigate Catastrophic Forgetting in Multimodal Large Language Models Model-Dowser：一种数据无关的重要性探测方法，用于缓解多模态大语言模型中的灾难性遗忘 large language model multimodal
2 Alignment Drift in Multimodal LLMs: A Two-Phase, Longitudinal Evaluation of Harm Across Eight Model Releases 纵向评估多模态LLM安全性：揭示八个模型版本中的对齐漂移现象 large language model multimodal
3 OmniSIFT: Modality-Asymmetric Token Compression for Efficient Omni-modal Large Language Models OmniSIFT：面向高效Omni-LLM的模态非对称Token压缩框架 large language model multimodal
4 Focus-LIME: Surgical Interpretation of Long-Context Large Language Models via Proxy-Based Neighborhood Selection Focus-LIME：通过代理模型邻域选择实现长文本LLM的可解释性 large language model
5 Revisiting Prompt Sensitivity in Large Language Models for Text Classification: The Role of Prompt Underspecification 研究表明，大语言模型文本分类中Prompt敏感性部分源于Prompt欠规范问题。 large language model
6 Bi-directional Bias Attribution: Debiasing Large Language Models without Modifying Prompts 提出双向偏差归因方法，无需修改提示即可消除大型语言模型中的偏见。 large language model ✅
7 DeFrame: Debiasing Large Language Models Against Framing Effects DeFrame：通过消除框架效应来提升大型语言模型的公平性 large language model
8 Inference-Time Reasoning Selectively Reduces Implicit Social Bias in Large Language Models 推理时推理选择性减少大型语言模型中的隐性社会偏见 large language model
9 Beyond Unimodal Shortcuts: MLLMs as Cross-Modal Reasoners for Grounded Named Entity Recognition 提出Modality-aware Consistency Reasoning (MCR)以解决GMNER中MLLM的模态偏见问题 large language model multimodal visual grounding
10 Evaluating the Presence of Sex Bias in Clinical Reasoning by Large Language Models 评估大型语言模型在临床推理中存在的性别偏见 large language model
11 History-Guided Iterative Visual Reasoning with Self-Correction 提出H-GIVR框架，通过历史信息引导迭代视觉推理并进行自校正，提升多模态大语言模型的推理可靠性。 large language model multimodal
12 CoT is Not the Chain of Truth: An Empirical Internal Analysis of Reasoning LLMs for Fake News Generation 揭示推理LLM生成虚假新闻时CoT的潜在风险，即使拒绝请求也可能包含不安全叙事。 large language model chain-of-thought
13 LinGO: A Linguistic Graph Optimization Framework with LLMs for Interpreting Intents of Online Uncivil Discourse LinGO：利用语言图优化框架与LLM提升在线不文明言论意图识别 large language model chain-of-thought
14 Can Vision Replace Text in Working Memory? Evidence from Spatial n-Back in Vision-Language Models 探究视觉信息在视觉-语言模型工作记忆中的作用：基于空间n-back任务的证据 large language model multimodal
15 Contextual Drag: How Errors in the Context Affect LLM Reasoning 揭示上下文拖拽效应：上下文错误如何影响大语言模型推理 large language model
16 Exploiting contextual information to improve stance detection in informal political discourse with LLMs 利用上下文信息，通过大语言模型提升非正式政治语境下的立场检测 large language model
17 VILLAIN at AVerImaTeC: Verifying Image-Text Claims via Multi-Agent Collaboration VILLAIN：基于多智能体协作验证图像-文本声明的系统 multimodal ✅
18 LycheeDecode: Accelerating Long-Context LLM Inference via Hybrid-Head Sparse Decoding LycheeDecode：通过混合头稀疏解码加速长文本LLM推理。 large language model
19 How Few-shot Demonstrations Affect Prompt-based Defenses Against LLM Jailbreak Attacks 探究Few-shot示例对Prompt防御LLM越狱攻击的影响，揭示RoP与ToP的差异。 large language model
20 Enforcing Monotonic Progress in Legal Cross-Examination: Preventing Long-Horizon Stagnation in LLM-Based Inquiry 提出Soft-FSM，通过外部状态控制解决LLM在法律交叉询问中长期停滞问题 large language model
21 Horizon-LM: A RAM-Centric Architecture for LLM Training Horizon-LM：一种以内存为中心的LLM训练架构，突破GPU限制。 large language model
22 Can LLMs capture stable human-generated sentence entropy measures? 研究表明LLM在多大程度上能捕捉人类句子熵的稳定性，并提供人类数据规范化的实践指南。 large language model
23 Fine-Grained Activation Steering: Steering Less, Achieving More AUSteer：通过细粒度激活控制，以更少干预实现更优大语言模型行为调控 large language model
24 Beyond Rejection Sampling: Trajectory Fusion for Scaling Mathematical Reasoning TrajFusion：通过轨迹融合提升LLM在数学推理中的性能 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🔗⭐
25 Guided Verifier: Collaborative Multimodal Reasoning via Dynamic Process Supervision 提出Guided Verifier框架，通过动态过程监督提升多模态大语言模型的推理能力。 reinforcement learning large language model multimodal
26 Reinforced Attention Learning 提出强化注意力学习RAL，直接优化多模态大语言模型的内部注意力分布，提升感知能力。 reinforcement learning distillation large language model
27 When Silence Is Golden: Can LLMs Learn to Abstain in Temporal QA and Beyond? 提出基于强化学习的CoT框架，提升LLM在时序问答中的拒答能力 reinforcement learning large language model chain-of-thought
28 ERNIE 5.0 Technical Report ERNIE 5.0：首个公开的万亿参数统一多模态自回归模型，支持理解与生成 reinforcement learning foundation model multimodal
29 Semantic Self-Distillation for Language Model Uncertainty 提出语义自蒸馏方法，用于语言模型不确定性量化和幻觉预测。 distillation large language model
30 ECG-R1: Protocol-Guided and Modality-Agnostic MLLM for Reliable ECG Interpretation 提出ECG-R1以解决ECG解读不可靠的问题 reinforcement learning large language model multimodal ✅
31 CoLT: Reasoning with Chain of Latent Tool Calls 提出CoLT框架，通过链式潜在工具调用提升LLM推理效率与精度 reinforcement learning large language model chain-of-thought
32 Language Models Struggle to Use Representations Learned In-Context 大型语言模型难以有效利用上下文学习到的表征完成下游任务 world model representation learning large language model
33 PersoDPO: Scalable Preference Optimization for Instruction-Adherent, Persona-Grounded Dialogue via Multi-LLM Evaluation PersoDPO：通过多LLM评估实现可扩展的、符合指令和角色设定的对话偏好优化 DPO direct preference optimization large language model
34 Scaling Agentic Verifier for Competitive Coding 提出Agentic Verifier，通过主动测试用例生成提升代码竞赛问题求解能力 reinforcement learning large language model

⬅️ 返回 cs.CL 首页 · 🏠 返回主页