cs.CL（2026-04-21）

📊 共 37 篇论文 | 🔗 10 篇有代码

🎯 兴趣领域导航

支柱九：具身大模型 (Embodied Foundation Models) (30 🔗6) 支柱二：RL算法与架构 (RL & Architecture) (5 🔗2) 支柱四：生成式动作 (Generative Motion) (1 🔗1) 支柱一：机器人控制 (Robot Control) (1 🔗1)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (30 篇)

#题目一句话要点标签🔗⭐
1 $R^2$-dLLM: Accelerating Diffusion Large Language Models via Spatio-Temporal Redundancy Reduction 提出$R^2$-dLLM以解决扩散大语言模型解码冗余问题 large language model
2 Cross-Model Consistency of AI-Generated Exercise Prescriptions: A Repeated Generation Study Across Three Large Language Models 评估大型语言模型在重复生成运动处方时的跨模型一致性 large language model
3 Impact of large language models on peer review opinions from a fine-grained perspective: Evidence from top conference proceedings in AI 研究表明大型语言模型影响同行评审意见，导致评审关注点从深度评估转向表面清晰度。 large language model
4 IndiaFinBench: An Evaluation Benchmark for Large Language Model Performance on Indian Financial Regulatory Text IndiaFinBench：首个面向印度金融监管文本的大语言模型评测基准 large language model ✅
5 Assessing Capabilities of Large Language Models in Social Media Analytics: A Multi-task Quest 全面评估大型语言模型在社交媒体分析三大任务中的能力，并建立可复现的基准。 large language model
6 Are Large Language Models Economically Viable for Industry Deployment? 提出EDGE-EVAL，弥合LLM工业部署评估中经济性与效率差距 large language model
7 Do Emotions Influence Moral Judgment in Large Language Models? 研究表明：情感会影响大语言模型的道德判断，且模型能力越强影响越小 large language model
8 AlignCultura: Towards Culturally Aligned Large Language Models? 提出AlignCultura，旨在提升大语言模型在文化维度上的对齐能力 large language model
9 Taming Actor-Observer Asymmetry in Agents via Dialectical Alignment 提出ReTAS，通过辩证对齐解决多智能体系统中行动者-观察者不对称性问题 large language model chain-of-thought
10 Chat2Workflow: A Benchmark for Generating Executable Visual Workflows with Natural Language 提出Chat2Workflow基准，用于评估大语言模型生成可执行可视化工作流的能力。 large language model ✅
11 LePREC: Reasoning as Classification over Structured Factors for Assessing Relevance of Legal Issues 提出LePREC框架，通过结构化推理提升LLM在法律问题相关性评估中的精度。 large language model
12 Rethinking Scale: Deployment Trade-offs of Small Language Models under Agent Paradigms 探索Agent范式下小型语言模型的部署权衡，提升资源受限场景性能 large language model
13 Location Not Found: Exposing Implicit Local and Global Biases in Multilingual LLMs LocQA揭示多语言LLM中隐含的地域和全局偏见 large language model
14 CulturALL: Benchmarking Multilingual and Multicultural Competence of LLMs on Grounded Tasks CulturALL：提出一个多语言文化常识基准，用于评估LLM在真实场景中的能力。 large language model
15 ShadowPEFT: Shadow Network for Parameter-Efficient Fine-Tuning 提出ShadowPEFT，通过深度共享的影子网络实现参数高效的LLM微调。 large language model
16 SAMoRA: Semantic-Aware Mixture of LoRA Experts for Task-Adaptive Learning SAMoRA：提出语义感知的LoRA专家混合模型，用于任务自适应学习 large language model ✅
17 Discovering a Shared Logical Subspace: Steering LLM Logical Reasoning via Alignment of Natural-Language and Symbolic Views 提出一种基于对齐自然语言和符号视图的LLM逻辑推理方法 large language model
18 Epistemic orientation in parliamentary discourse is associated with deliberative democracy 提出基于LLM的EMI指标，评估议会言论的认知倾向与审议民主的关联性 large language model
19 Bangla Key2Text: Text Generation from Keywords for a Low Resource Language Bangla Key2Text：为低资源语言孟加拉语构建关键词到文本生成的大规模数据集。 large language model
20 Beyond Rating: A Comprehensive Evaluation and Benchmark for AI Reviews 提出Beyond Rating框架，从文本论证角度全面评估AI评审质量，解决现有benchmark仅关注评分预测的局限性。 large language model
21 What Makes an LLM a Good Optimizer? A Trajectory Analysis of LLM-Guided Evolutionary Search 通过轨迹分析揭示LLM作为优化器的内在机制，助力进化搜索系统设计 large language model
22 Lost in Translation: Do LVLM Judges Generalize Across Languages? 提出MM-JudgeBench，评估LVLM评判模型在多语言环境下的泛化能力。 multimodal
23 DASH-KV: Accelerating Long-Context LLM Inference via Asymmetric KV Cache Hashing DASH-KV：通过非对称KV缓存哈希加速长文本LLM推理 large language model ✅
24 Evaluating LLM-Driven Summarisation of Parliamentary Debates with Computational Argumentation 提出基于计算论证的框架，评估LLM生成的议会辩论摘要在论证内容上的忠实性。 large language model
25 HarDBench: A Benchmark for Draft-Based Co-Authoring Jailbreak Attacks for Safe Human-LLM Collaborative Writing HarDBench：提出基于草稿的协同写作越狱攻击评测基准，保障人机协作安全 large language model ✅
26 Headlines You Won't Forget: Can Pronoun Insertion Increase Memorability? 研究人称代词插入对新闻标题记忆效果的影响，并评估LLM自动插入的质量。 large language model
27 Mind the Unseen Mass: Unmasking LLM Hallucinations via Soft-Hybrid Alphabet Estimation 提出SHADE，通过软混合字母表估计解决LLM在小样本下的幻觉问题。 large language model
28 Construction of Knowledge Graph based on Language Model 提出基于轻量级语言模型的超关系知识图谱构建框架 large language model
29 Detoxification for LLM: From Dataset Itself 提出HSPD框架，通过数据清洗从源头减少LLM的毒性，提升模型安全性。 large language model ✅
30 HoWToBench: Holistic Evaluation for LLM's Capability in Human-level Writing using Tree of Writing 提出HoWToBench基准与Tree-of-Writing评估方法，解决LLM写作能力评估难题。 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🔗⭐
31 TRN-R1-Zero: Text-rich Network Reasoning via LLMs with Reinforcement Learning Only 提出TRN-R1-Zero以解决文本丰富网络推理问题 reinforcement learning distillation large language model ✅
32 The Rise of Verbal Tics in Large Language Models: A Systematic Analysis Across Frontier Models 系统分析前沿大语言模型中的语言习惯性重复现象，揭示对齐训练的潜在代价。 reinforcement learning RLHF large language model
33 Deep Supervised Contrastive Learning of Pitch Contours for Robust Pitch Accent Classification in Seoul Korean 提出Dual-Glob框架，利用深度监督对比学习提升首尔韩语音高重音分类的鲁棒性 predictive model contrastive learning
34 SCURank: Ranking Multiple Candidate Summaries with Summary Content Units for Enhanced Summarization 提出SCURank，利用摘要内容单元提升多候选摘要排序质量，增强文本摘要能力 distillation large language model ✅
35 Pause or Fabricate? Training Language Models for Grounded Reasoning 提出GRIL框架，解决大语言模型在不完备信息下的无根据推理问题 reinforcement learning large language model

🔬 支柱四：生成式动作 (Generative Motion) (1 篇)

#题目一句话要点标签🔗⭐
36 Emotion-Cause Pair Extraction in Conversations via Semantic Decoupling and Graph Alignment 提出SCALE框架，通过解耦语义和图对齐解决对话中的情感原因对抽取问题。 motion diffusion ✅

🔬 支柱一：机器人控制 (Robot Control) (1 篇)

#题目一句话要点标签🔗⭐
37 Debating the Unspoken: Role-Anchored Multi-Agent Reasoning for Half-Truth Detection 提出RADAR框架，通过角色扮演多智能体辩论检测半真陈述，提升事实核查的准确性。 manipulation ✅

⬅️ 返回 cs.CL 首页 · 🏠 返回主页