cs.CL（2025-07-24）

📊 共 36 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) (31 🔗6) 支柱二：RL算法与架构 (RL & Architecture) (5)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (31 篇)

#题目一句话要点标签🔗⭐
1 How does Chain of Thought Think? Mechanistic Interpretability of Chain-of-Thought Reasoning with Sparse Autoencoding 通过稀疏自编码器，研究思维链（CoT）推理的机制可解释性 large language model chain-of-thought
2 FeynTune: Large Language Models for High-Energy Theory FeynTune：利用大型语言模型进行高能理论研究的专用模型微调 large language model
3 The Moral Gap of Large Language Models 揭示大型语言模型在道德推理上的局限性，表明微调模型优于提示工程。 large language model
4 Evaluating Large Language Models (LLMs) in Financial NLP: A Comparative Study on Financial Report Analysis 对比评估大型语言模型在金融报告分析中的表现，揭示GPT模型的优越性。 large language model
5 Deep Learning Approaches for Multimodal Intent Recognition: A Survey 综述深度学习在多模态意图识别中的应用，分析方法、数据集、挑战与未来方向。 multimodal
6 GIIFT: Graph-guided Inductive Image-free Multimodal Machine Translation 提出GIIFT框架，利用图结构引导的归纳式无图多模态机器翻译，显著提升翻译效果。 multimodal
7 StyleAdaptedLM: Enhancing Instruction Following Models with Efficient Stylistic Transfer StyleAdaptedLM：利用高效风格迁移增强指令跟随模型 instruction following
8 SCOPE: Stochastic and Counterbiased Option Placement for Evaluating Large Language Models SCOPE：通过随机和反偏置选项放置评估大语言模型 large language model
9 HIVMedQA: Benchmarking large language models for HIV medical decision support HIVMedQA：评估大型语言模型在HIV医疗决策支持中的性能 large language model
10 Hybrid and Unitary PEFT for Resource-Efficient Large Language Models 提出混合正交稳定与梯度对齐PEFT方法，高效微调大规模语言模型 large language model
11 Synthetic Data Generation for Phrase Break Prediction with Large Language Model 利用大型语言模型生成合成数据，解决短语停顿预测的数据标注难题 large language model
12 EH-Benchmark Ophthalmic Hallucination Benchmark and Agent-Driven Top-Down Traceable Reasoning Workflow 提出EH-Benchmark眼科幻觉基准及Agent驱动的可溯源推理工作流，提升眼科诊断准确性。 large language model multimodal ✅
13 GrAInS: Gradient-based Attribution for Inference-Time Steering of LLMs and VLMs GrAInS：利用梯度归因实现LLM和VLM的推理时引导 large language model multimodal
14 BadReasoner: Planting Tunable Overthinking Backdoors into Large Reasoning Models for Fun or Profit 提出BadReasoner，通过可控的过度推理后门攻击大型推理模型 large language model chain-of-thought ✅
15 TRPrompt: Bootstrapping Query-Aware Prompt Optimization from Textual Rewards TRPrompt：从文本奖励引导查询感知的提示优化，提升LLM推理能力 large language model
16 Not All Features Deserve Attention: Graph-Guided Dependency Learning for Tabular Data Generation with Language Models 提出GraDe，利用图引导依赖学习提升语言模型在表格数据生成中的性能。 large language model
17 CLEAR: Error Analysis via LLM-as-a-Judge Made Easy CLEAR：利用LLM作为评判者进行简易的错误分析 large language model
18 Trusted Knowledge Extraction for Operations and Maintenance Intelligence 针对航空运维情报，提出可信知识抽取方法，解决数据保密与集成难题。 large language model
19 AQuilt: Weaving Logic and Self-Inspection into Low-Cost, High-Relevance Data Synthesis for Specialist LLMs AQuilt：通过逻辑推理与自检，低成本合成高质量领域专家LLM训练数据 large language model ✅
20 GLiNER2: An Efficient Multi-Task Information Extraction System with Schema-Driven Interface 提出GLiNER2以解决信息提取任务的多样性与效率问题 large language model ✅
21 Enhancing RAG Efficiency with Adaptive Context Compression 提出ACC-RAG，通过自适应上下文压缩提升RAG效率 large language model
22 CodeMixBench: Evaluating Code-Mixing Capabilities of LLMs Across 18 Languages CodeMixBench：提出多语言混合代码能力评测基准，揭示LLM在跨语系混合场景下的性能瓶颈。 large language model ✅
23 Wide-In, Narrow-Out: Revokable Decoding for Efficient and Effective DLLMs 提出WINO算法，通过可撤销解码显著提升扩散大语言模型的速度与质量。 large language model
24 Protecting Vulnerable Voices: Synthetic Dataset Generation for Self-Disclosure Detection 提出一种基于LLM的合成数据集生成方法，用于保护社交平台中易受攻击群体的个人信息自披露行为。 large language model
25 Assemble Your Crew: Automatic Multi-agent Communication Topology Design via Autoregressive Graph Generation 提出ARG-Designer，通过自回归图生成自动设计多智能体通信拓扑，提升任务适应性。 large language model ✅
26 Exploring the Impact of Instruction-Tuning on LLM's Susceptibility to Misinformation 研究表明指令微调提升LLM可用性的同时，显著增加其对虚假信息的接受度 large language model
27 Safeguarding RAG Pipelines with GMTP: A Gradient-based Masked Token Probability Method for Poisoned Document Detection 提出GMTP方法，通过梯度分析和掩码概率检测RAG管道中的恶意文档。 large language model
28 MathOPEval: A Fine-grained Evaluation Benchmark for Visual Operations of MLLMs in Mathematical Reasoning MathOPEval：一个用于评估MLLM在数学推理中视觉操作能力的细粒度基准 large language model
29 Privacy-Preserving Synthetic Review Generation with Diverse Writing Styles Using LLMs 提出基于提示的方法，提升LLM生成合成评论的多样性并保护隐私 large language model
30 Resource Consumption Red-Teaming for Large Vision-Language Models 提出RECITE，通过视觉引导优化实现LVLM的资源消耗红队测试。 large language model
31 NeuralDB: Scaling Knowledge Editing in LLMs to 100,000 Facts with Neural KV Database 提出NeuralDB以高效编辑大规模知识库中的事实 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🔗⭐
32 FinMarBa: A Market-Informed Dataset for Financial Sentiment Classification 提出结合金融新闻情感与市场指标的分层强化学习投资组合优化框架 reinforcement learning deep reinforcement learning DRL
33 Technical Report of TeleChat2, TeleChat2.5 and T1 TeleChat2系列模型：通过增强训练策略显著提升性能，并在复杂推理任务上超越GPT-4o reinforcement learning DPO direct preference optimization
34 FinDPO: Financial Sentiment Analysis for Algorithmic Trading through Preference Optimization of LLMs FinDPO：通过偏好优化LLM进行金融情感分析，提升算法交易收益。 DPO direct preference optimization large language model
35 Checklists Are Better Than Reward Models For Aligning Language Models 提出基于检查清单反馈的强化学习(RLCF)，提升语言模型指令遵循能力。 reinforcement learning instruction following
36 Hybrid Annotation for Propaganda Detection: Integrating LLM Pre-Annotations with Human Intelligence 提出一种混合标注框架，结合LLM预标注与人工审核，提升宣传检测的效率与一致性。 distillation large language model

⬅️ 返回 cs.CL 首页 · 🏠 返回主页