cs.CL（2026-04-09）

📊 共 32 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) (6 🔗1) 支柱六：视频提取与匹配 (Video Extraction) (1) 支柱一：机器人控制 (Robot Control) (1)

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

#题目一句话要点标签🔗⭐
1 SeLaR: Selective Latent Reasoning in Large Language Models 提出SeLaR，通过选择性隐空间推理提升大语言模型的推理能力。 large language model chain-of-thought
2 Self-Debias: Self-correcting for Debiasing Large Language Models Self-Debias：通过自校正机制消除大语言模型中的偏见传播 large language model chain-of-thought
3 Rethinking Data Mixing from the Perspective of Large Language Models 提出DoGraph框架，通过图约束优化重加权数据，提升大语言模型泛化能力 large language model
4 GRASS: Gradient-based Adaptive Layer-wise Importance Sampling for Memory-efficient Large Language Model Fine-tuning GRASS：基于梯度自适应层重要性采样，实现大语言模型高效微调。 large language model
5 Distributed Multi-Layer Editing for Rule-Level Knowledge in Large Language Models 提出分布式多层编辑(DMLE)方法，解决大语言模型中规则级知识编辑难题。 large language model ✅
6 Detecting HIV-Related Stigma in Clinical Narratives Using Large Language Models 利用大型语言模型检测临床叙述中与HIV相关的污名化现象 large language model
7 Towards Real-world Human Behavior Simulation: Benchmarking Large Language Models on Long-horizon, Cross-scenario, Heterogeneous Behavior Traces OmniBehavior：构建真实世界人类行为模拟基准，揭示LLM在复杂行为建模中的局限性 large language model
8 Beyond Social Pressure: Benchmarking Epistemic Attack in Large Language Models 提出PPT-Bench基准，用于评估大语言模型在认知攻击下的脆弱性 large language model
9 Are GUI Agents Focused Enough? Automated Distraction via Semantic-level UI Element Injection 提出语义级UI元素注入方法，用于评估GUI智能体的鲁棒性并发现潜在安全漏洞。 visual grounding
10 Loop, Think, & Generalize: Implicit Reasoning in Recurrent-Depth Transformers 提出循环深度Transformer，解决Transformer在隐式推理中组合泛化能力不足的问题。 large language model
11 A GAN and LLM-Driven Data Augmentation Framework for Dynamic Linguistic Pattern Modeling in Chinese Sarcasm Detection 提出基于GAN和LLM的数据增强框架，用于动态建模中文讽刺检测中的语言模式。 large language model
12 TEMPER: Testing Emotional Perturbation in Quantitative Reasoning TEMPER：探究情感扰动对定量推理的影响及中和方法 large language model
13 Cram Less to Fit More: Training Data Pruning Improves Memorization of Facts 提出基于训练损失的数据剪枝方法，提升大语言模型的事实记忆能力 large language model
14 What do Language Models Learn and When? The Implicit Curriculum Hypothesis 揭示LLM预训练的隐式课程：技能以可预测的组合方式涌现 large language model
15 AI generates well-liked but templatic empathic responses 大型语言模型生成受欢迎但模板化的共情回复 large language model
16 Training Data Size Sensitivity in Unsupervised Rhyme Recognition 研究揭示了无监督韵律识别中训练数据规模对性能的影响，并提出了RhymeTagger。 large language model
17 HCRE: LLM-based Hierarchical Classification for Cross-Document Relation Extraction with a Prediction-then-Verification Strategy 提出基于LLM的分层分类模型HCRE，解决跨文档关系抽取中关系数量过多的挑战。 large language model
18 Tool Retrieval Bridge: Aligning Vague Instructions with Retriever Preferences via Bridge Model 提出工具检索桥TRB，解决LLM在模糊指令下的工具检索问题 large language model ✅
19 An Empirical Analysis of Static Analysis Methods for Detection and Mitigation of Code Library Hallucinations 利用静态分析检测和缓解代码库幻觉问题，揭示其能力上限。 large language model
20 SepSeq: A Training-Free Framework for Long Numerical Sequence Processing in LLMs SepSeq：一种免训练框架，通过分隔符提升LLM长数值序列处理能力 large language model
21 LLMs Underperform Graph-Based Parsers on Supervised Relation Extraction for Complex Graphs 复杂图关系抽取中，图解析器性能优于大型语言模型 large language model
22 Optimal Multi-bit Generative Watermarking Schemes Under Worst-Case False-Alarm Constraints 针对大语言模型，提出在最坏情况误报约束下的最优多比特生成式水印方案 large language model
23 EXAONE 4.5 Technical Report LG AI Research发布EXAONE 4.5，首个开源权重视觉语言模型，提升文档理解与长文本推理能力。 multimodal
24 An Empirical Analysis of Static Analysis Methods for Detection and Mitigation of Code Library Hallucinations 利用静态分析检测和缓解代码库幻觉问题，揭示其能力上限 large language model

🔬 支柱二：RL算法与架构 (RL & Architecture) (6 篇)

#题目一句话要点标签🔗⭐
25 Large Language Model Post-Training: A Unified View of Off-Policy and On-Policy Learning 统一视角解读大语言模型后训练：离线与在线学习的融合 reinforcement learning policy learning distillation
26 Demystifying OPD: Length Inflation and Stabilization Strategies for Large Language Models 提出StableOPD，解决On-policy蒸馏中长度膨胀和训练不稳定的问题 distillation large language model
27 TSUBASA: Improving Long-Horizon Personalization via Evolving Memory and Self-Learning with Context Distillation TSUBASA：通过动态记忆演化和上下文蒸馏自学习，提升长程个性化语言模型能力 distillation large language model
28 The Art of (Mis)alignment: How Fine-Tuning Methods Effectively Misalign and Realign LLMs in Post-Training 提出细调方法以解决大型语言模型的对齐与失调问题 DPO direct preference optimization large language model ✅
29 A Decomposition Perspective to Long-context Reasoning for LLMs 提出长文本推理分解方法，通过强化学习提升LLM在原子技能上的表现，进而增强长文本推理能力。 reinforcement learning large language model
30 Guaranteeing Knowledge Integration with Joint Decoding for Retrieval-Augmented Generation GuarantRAG：通过联合解码保证知识整合的检索增强生成框架 DPO large language model

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

#题目一句话要点标签🔗⭐
31 Cards Against LLMs: Benchmarking Humor Alignment in Large Language Models 评估LLM幽默感：使用“反对LLM的卡牌”基准测试模型与人类幽默偏好的一致性 HuMoR large language model

🔬 支柱一：机器人控制 (Robot Control) (1 篇)

#题目一句话要点标签🔗⭐
32 MT-OSC: Path for LLMs that Get Lost in Multi-Turn Conversation MT-OSC：解决LLM在多轮对话中迷失问题的路径，实现高效上下文压缩。 OSC large language model

⬅️ 返回 cs.CL 首页 · 🏠 返回主页