cs.CL（2025-09-25）

📊 共 35 篇论文 | 🔗 6 篇有代码

🎯 兴趣领域导航

支柱九：具身大模型 (Embodied Foundation Models) (25 🔗4) 支柱二：RL算法与架构 (RL & Architecture) (9 🔗2) 支柱三：空间感知与语义 (Perception & Semantics) (1)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (25 篇)

#题目一句话要点标签🔗⭐
1 ReviewScore: Misinformed Peer Review Detection with Large Language Models 提出ReviewScore以检测同行评审中的错误信息，提升评审质量。 large language model
2 Bounds of Chain-of-Thought Robustness: Reasoning Steps, Embed Norms, and Beyond 理论分析CoT推理的鲁棒性边界，揭示推理步数和嵌入范数的影响 chain-of-thought
3 CLaw: Benchmarking Chinese Legal Knowledge in Large Language Models - A Fine-grained Corpus and Reasoning Analysis CLaw：构建中文法律知识基准，评估大语言模型在法律推理中的能力。 large language model
4 BESPOKE: Benchmark for Search-Augmented Large Language Model Personalization via Diagnostic Feedback 提出BESPOKE基准，用于诊断反馈驱动的搜索增强LLM个性化 large language model ✅
5 PerHalluEval: Persian Hallucination Evaluation Benchmark for Large Language Models 提出PerHalluEval，首个波斯语LLM幻觉评估基准 large language model
6 SoM-1K: A Thousand-Problem Benchmark Dataset for Strength of Materials 提出SoM-1K材料力学基准数据集，评估并提升多模态工程问题中大模型的性能。 large language model foundation model multimodal
7 LLM-Based Support for Diabetes Diagnosis: Opportunities, Scenarios, and Challenges with GPT-5 利用GPT-5辅助糖尿病诊断，提升临床决策支持与患者理解 large language model multimodal
8 When Instructions Multiply: Measuring and Estimating LLM Capabilities of Multiple Instructions Following 提出ManyIFEval和StyleMBPP基准，评估并预测LLM多指令遵循能力。 large language model instruction following
9 OjaKV: Context-Aware Online Low-Rank KV Cache Compression with Oja's Rule OjaKV：利用Oja规则进行上下文感知在线低秩KV缓存压缩，提升长文本处理效率。 large language model
10 Hallucination-Resistant, Domain-Specific Research Assistant with Self-Evaluation and Vector-Grounded Retrieval 提出RA-FSM，一种抗幻觉、领域特定的研究助手，通过自评估和向量检索提升专家工作流效率。 large language model
11 Generation-Time vs. Post-hoc Citation: A Holistic Evaluation of LLM Attribution 对比生成时和后置引用，全面评估LLM的归因能力，为高风险场景提供选择依据。 large language model
12 On Code-Induced Reasoning in LLMs 系统性研究代码特性对LLM推理能力的影响，揭示结构与语义扰动的关键作用 large language model
13 One Model, Many Morals: Uncovering Cross-Linguistic Misalignments in Computational Moral Reasoning 揭示多语言环境下大语言模型道德推理的跨语言错位问题 large language model
14 Sycophancy Is Not One Thing: Causal Separation of Sycophantic Behaviors in LLMs 因LLM溜须拍马行为并非单一机制，论文提出因果分离方法以独立控制不同行为。 large language model
15 The role of synthetic data in Multilingual, Multi-cultural AI systems: Lessons from Indic Languages 提出Updesh数据集，利用合成数据提升多语言、多文化AI系统在印度语言上的性能。 instruction following
16 DisCoCLIP: A Distributional Compositional Tensor Network Encoder for Vision-Language Understanding DisCoCLIP：一种用于视觉-语言理解的分布组合张量网络编码器 multimodal
17 LLMTrace: A Corpus for Classification and Fine-Grained Localization of AI-Written Text LLMTrace：用于AI生成文本分类与精细定位的双语数据集 large language model ✅
18 LLM Output Homogenization is Task Dependent 提出任务依赖的LLM输出同质化评估与缓解方法，提升功能多样性。 large language model
19 Eigen-1: Adaptive Multi-Agent Refinement with Monitor-Based RAG for Scientific Reasoning Eigen-1：基于Monitor的RAG自适应多智能体精炼，用于科学推理 large language model ✅
20 Who's Laughing Now? An Overview of Computational Humour Generation and Explanation 计算幽默生成与解释综述：探索NLP在幽默理解与创造中的应用与挑战 large language model
21 Which Cultural Lens Do Models Adopt? On Cultural Positioning Bias and Agentic Mitigation in LLMs 揭示LLM文化定位偏差并提出基于Agent的偏见缓解方法 large language model
22 PMark: Towards Robust and Distortion-free Semantic-level Watermarking with Channel Constraints PMark：基于通道约束的鲁棒无失真语义级水印方法 large language model ✅
23 Behind RoPE: How Does Causal Mask Encode Positional Information? 揭示RoPE背后机制：因果掩码如何编码位置信息 large language model
24 Generative AI for FFRDCs 利用生成式AI加速FFRDC文本分析，提升政府机构效率与安全性 large language model
25 Analysis of instruction-based LLMs' capabilities to score and judge text-input problems in an academic setting 提出基于LLM的自动评分系统，用于评估学术文本输入问题，参考答案辅助效果最佳。 large language model

🔬 支柱二：RL算法与架构 (RL & Architecture) (9 篇)

#题目一句话要点标签🔗⭐
26 Multi-Objective Reinforcement Learning for Large Language Model Optimization: Visionary Perspective 针对大语言模型优化，提出多目标强化学习的远景框架 reinforcement learning large language model
27 Retrieval over Classification: Integrating Relation Semantics for Multimodal Relation Extraction 提出ROC框架，将多模态关系抽取重构为检索任务，提升细粒度关系理解能力。 contrastive learning large language model multimodal
28 Painless Activation Steering: An Automated, Lightweight Approach for Post-Training Large Language Models 提出Painless Activation Steering (PAS)，一种全自动、轻量级的后训练大语言模型激活向量调控方法。 reinforcement learning large language model
29 SciReasoner: Laying the Scientific Reasoning Ground Across Disciplines SciReasoner：构建跨学科的科学推理基础模型 reinforcement learning reward shaping foundation model ✅
30 Dual-Head Reasoning Distillation: Improving Classifier Accuracy with Train-Time-Only Reasoning 提出双头推理蒸馏(DHRD)，在不牺牲推理速度的前提下提升分类器精度。 distillation chain-of-thought
31 Learning to Reason with Mixture of Tokens 提出混合Token生成方法，提升LLM在可验证奖励强化学习中的推理能力。 reinforcement learning large language model chain-of-thought
32 Hallucination reduction with CASAL: Contrastive Activation Steering For Amortized Learning CASAL：对比激活引导的摊销学习，有效降低大语言模型幻觉 DPO large language model
33 A State-of-the-Art SQL Reasoning Model using RLVR 利用可验证奖励的强化学习，提出SQL推理模型RLVR，在BIRD数据集上达到SOTA。 reinforcement learning offline RL
34 RLBFF: Binary Flexible Feedback to bridge between Human Feedback & Verifiable Rewards 提出RLBFF，结合人类反馈和可验证奖励，提升LLM对齐效果并支持推理时自定义原则。 reinforcement learning RLHF ✅

🔬 支柱三：空间感知与语义 (Perception & Semantics) (1 篇)

#题目一句话要点标签🔗⭐
35 Vision Language Models Cannot Plan, but Can They Formalize? 提出VLM作为形式化工具以解决多模态规划问题 open-vocabulary open vocabulary multimodal

⬅️ 返回 cs.CL 首页 · 🏠 返回主页