cs.CL（2025-02-11）

📊 共 33 篇论文 | 🔗 9 篇有代码

🎯 兴趣领域导航

支柱九：具身大模型 (Embodied Foundation Models) (24 🔗4) 支柱二：RL算法与架构 (RL & Architecture) (8 🔗4) 支柱六：视频提取与匹配 (Video Extraction) (1 🔗1)

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

#题目一句话要点标签🔗⭐
1 BenchMAX: A Comprehensive Multilingual Evaluation Suite for Large Language Models BenchMAX：一个全面的多语言评估套件，用于评估大型语言模型 large language model instruction following
2 GENERator: A Long-Context Generative Genomic Foundation Model 提出GENERator：一个长上下文的生成式基因组基础模型 large language model foundation model ✅
3 AI-VERDE: A Gateway for Egalitarian Access to Large Language Model-Based Resources For Educational Institutions AI-VERDE：为教育机构提供平等的LLM资源访问平台 large language model
4 Large Language Models as Proxies for Theories of Human Linguistic Cognition 利用大型语言模型作为人类语言认知理论的代理模型 large language model
5 Exploring Mobile Touch Interaction with Large Language Models 提出一种基于触摸手势的LLM交互方法，用于移动设备上的文本编辑。 large language model
6 Target-Augmented Shared Fusion-based Multimodal Sarcasm Explanation Generation 提出TURBO模型，通过目标增强的共享融合机制提升多模态讽刺解释生成效果。 multimodal
7 Bridging the Evaluation Gap: Leveraging Large Language Models for Topic Model Evaluation 利用大型语言模型自动评估科学文献中动态演化的主题模型 large language model
8 Aligning Large Language Models to Follow Instructions and Hallucinate Less via Effective Data Filtering 提出NOVA框架，通过数据过滤减少大语言模型指令调优中的幻觉问题 large language model
9 CodeI/O: Condensing Reasoning Patterns via Code Input-Output Prediction 提出CodeI/O，通过代码输入输出预测提炼通用推理模式，提升大语言模型在多任务上的推理能力。 large language model chain-of-thought ✅
10 Speculate, then Collaborate: Fusing Knowledge of Language Models during Decoding 提出协同推测解码（CoSD），在解码时融合多个LLM知识，提升性能。 large language model
11 Hallucination, Monofacts, and Miscalibration: An Empirical Investigation 通过控制单因素率和选择性加权，实证研究LLM幻觉现象的成因与缓解策略 large language model
12 The Geometry of Prompting: Unveiling Distinct Mechanisms of Task Adaptation in Language Models 揭示Prompt几何特性：探究语言模型任务自适应的不同机制 large language model
13 Auditing Prompt Caching in Language Model APIs 通过时序审计揭示LLM API中的Prompt缓存及潜在隐私泄露风险 large language model
14 WHODUNIT: Evaluation benchmark for culprit detection in mystery stories 提出WhoDunIt数据集，评估LLM在推理小说中识别罪犯的演绎推理能力 large language model
15 Mask-Enhanced Autoregressive Prediction: Pay Less Attention to Learn More 提出Mask-Enhanced Autoregressive Prediction (MEAP)以增强LLM上下文检索能力 large language model
16 PerCul: A Story-Driven Cultural Evaluation of LLMs in Persian PerCul：提出波斯语文化评估数据集，用于评估LLM的文化敏感性。 large language model ✅
17 Forget What You Know about LLMs Evaluations -- LLMs are Like a Chameleon 提出C-BOD框架，用于检测大型语言模型在基准测试中的过拟合现象 large language model
18 RomanLens: The Role Of Latent Romanization In Multilinguality In LLMs RomanLens：揭示LLM多语言能力中潜在的罗马化作用 large language model
19 Entity Linking using LLMs for Automated Product Carbon Footprint Estimation 利用大型语言模型进行实体链接，实现产品碳足迹的自动评估。 large language model
20 MEMIT-Merge: Addressing MEMIT's Key-Value Conflicts in Same-Subject Batch Editing for LLMs MEMIT-Merge：解决MEMIT在LLM同主题批量编辑中的Key-Value冲突问题 large language model
21 Small Language Model Makes an Effective Long Text Extractor 提出SeNER，一种轻量级长文本实体抽取方法，显著提升抽取精度并降低内存占用。 large language model ✅
22 A Large-Scale Benchmark for Vietnamese Sentence Paraphrases 构建大规模高质量越南语句子释义数据集ViSP，促进越南语自然语言处理研究。 large language model
23 Does Training on Synthetic Data Make Models Less Robust? 研究表明，使用合成数据训练LLM并不会加剧其在NLI任务中的固有盲点。 large language model
24 Language-TPP: Integrating Temporal Point Processes with Language Models for Event Analysis Language-TPP：融合时序点过程与语言模型，用于增强事件分析能力。 large language model

🔬 支柱二：RL算法与架构 (RL & Architecture) (8 篇)

#题目一句话要点标签🔗⭐
25 LongReD: Mitigating Short-Text Degradation of Long-Context Large Language Models via Restoration Distillation LongReD：通过恢复蒸馏缓解长文本大模型短文本性能退化 distillation large language model ✅
26 Refine Knowledge of Large Language Models via Adaptive Contrastive Learning 提出自适应对比学习，提升大语言模型知识的准确性，缓解幻觉问题 contrastive learning large language model
27 DPO-Shift: Shifting the Distribution of Direct Preference Optimization 提出DPO-Shift以控制DPO的分布偏移，缓解语言模型训练中的likelihood displacement问题。 DPO direct preference optimization ✅
28 O1 Embedder: Let Retrievers Think Before Action O1 Embedder：让检索器在行动前先思考，提升复杂检索任务性能 behavior cloning contrastive learning large language model
29 Towards Efficient and Multifaceted Computer-assisted Pronunciation Training Leveraging Hierarchical Selective State Space Model and Decoupled Cross-entropy Loss 提出HMamba模型，融合APA和MDD任务，提升计算机辅助发音训练效率。 Mamba state space model ✅
30 Bridging Brain Signals and Language: A Deep Learning Approach to EEG-to-Text Decoding 提出一种基于深度学习的脑电信号到文本解码框架，实现开放词汇和个性化脑机接口。 representation learning open-vocabulary open vocabulary
31 Principled Data Selection for Alignment: The Hidden Risks of Difficult Examples 提出Selective DPO，通过数据选择提升大语言模型对齐效果 DPO large language model ✅
32 Adapting Multilingual Embedding Models to Historical Luxembourgish 针对历史卢森堡语，提出自适应多语言嵌入模型以提升跨语言语义搜索性能。 contrastive learning distillation

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

#题目一句话要点标签🔗⭐
33 Hierarchical Document Parsing via Large Margin Feature Matching and Heuristics 提出基于大间隔特征匹配和启发式的分层文档解析方法，在AAAI-25 VRD-IU挑战赛中获得第一名。 feature matching ✅

⬅️ 返回 cs.CL 首页 · 🏠 返回主页