cs.CL（2024-12-19）

📊 共 53 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) (42 🔗7) 支柱二：RL算法与架构 (RL & Architecture) (10 🔗2) 支柱一：机器人控制 (Robot Control) (1)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (42 篇)

#题目一句话要点标签🔗⭐
1 LMFusion: Adapting Pretrained Language Models for Multimodal Generation LMFusion：通过适配预训练语言模型实现多模态生成 large language model multimodal
2 Progressive Multimodal Reasoning via Active Retrieval 提出AR-MCTS框架，通过主动检索和蒙特卡洛树搜索提升多模态大语言模型的多步推理能力。 large language model multimodal
3 Conceptual In-Context Learning and Chain of Concepts: Solving Complex Conceptual Problems Using Large Language Models 提出概念内上下文学习与概念链，增强LLM解决复杂概念问题的能力 large language model chain-of-thought
4 PsyDraw: A Multi-Agent Multimodal System for Mental Health Screening in Left-Behind Children PsyDraw：一种多智能体多模态系统，用于留守儿童的心理健康筛查。 large language model multimodal
5 Adaptive Pruning for Large Language Models with Structural Importance Awareness 提出结构感知自适应剪枝方法SAAP，用于压缩LLM并在资源受限设备上部署。 large language model
6 A Comparative Study of DSPy Teleprompter Algorithms for Aligning Large Language Models Evaluation Metrics to Human Evaluation 对比DSPy Teleprompter算法，优化LLM提示以对齐人类评估标准 large language model
7 Confidence in the Reasoning of Large Language Models 评估大语言模型推理置信度：定性分析与量化指标相结合 large language model
8 Eliciting Causal Abilities in Large Language Models for Reasoning Tasks 提出SCIE方法，通过诱导大语言模型的因果推理能力提升其在推理任务中的表现。 large language model
9 RobustFT: Robust Supervised Fine-tuning for Large Language Models under Noisy Response 提出RobustFT框架，解决大语言模型在噪声响应下的鲁棒微调问题 large language model
10 Each Fake News is Fake in its Own Way: An Attribution Multi-Granularity Benchmark for Multimodal Fake News Detection 构建多粒度属性基准数据集AMG，并提出多粒度线索对齐模型MGCM，用于多模态假新闻检测与溯源。 multimodal
11 Analysis and Visualization of Linguistic Structures in Large Language Models: Neural Representations of Verb-Particle Constructions in BERT 分析大型语言模型中动词-小品词结构的神经表征：以BERT为例 large language model
12 ORBIT: Cost-Effective Dataset Curation for Large Language Model Domain Adaptation with an Astronomy Case Study ORBIT：一种低成本的大语言模型领域自适应数据集构建方法 large language model ✅
13 Why Do Large Language Models (LLMs) Struggle to Count Letters? 研究揭示大语言模型在字母计数任务上的困难，并分析其与词频、复杂度的关系 large language model
14 Graph-Convolutional Networks: Named Entity Recognition and Large Language Model Embedding in Document Clustering 提出一种融合命名实体识别和LLM嵌入的图卷积网络文档聚类方法 large language model
15 ResoFilter: Fine-grained Synthetic Data Filtering for Large Language Models through Data-Parameter Resonance Analysis ResoFilter：通过数据-参数共振分析实现大语言模型精细化合成数据过滤 large language model
16 A Large-Scale Simulation on Large Language Models for Decision-Making in Political Science 提出基于大语言模型的多步推理框架，用于大规模模拟政治决策 large language model
17 Sliding Windows Are Not the End: Exploring Full Ranking with Long-Context Large Language Models 提出长上下文大语言模型以解决滑动窗口策略的效率问题 large language model ✅
18 Why We Build Local Large Language Models: An Observational Analysis from 35 Japanese and Multilingual LLMs 通过观察性分析，揭示了构建本地大型语言模型的必要性与策略。 large language model
19 Are Longer Prompts Always Better? Prompt Selection in Large Language Models for Recommendation Systems 针对LLM推荐系统，提出基于数据集特征的Prompt选择方法，提升推荐准确率和效率。 large language model
20 Systematic Evaluation of Long-Context LLMs on Financial Concepts 系统性评估长文本LLM在金融概念理解上的能力，揭示其在长上下文中的脆弱性 large language model instruction following
21 Query pipeline optimization for cancer patient question answering systems 针对癌症患者问答系统，提出RAG查询管道三方面优化方法，提升回答准确率。 large language model chain-of-thought
22 Length Controlled Generation for Black-box LLMs 提出基于Metropolis-Hastings算法的迭代采样框架，实现黑盒LLM的精确长度控制 large language model instruction following
23 MMLU-CF: A Contamination-free Multi-task Language Understanding Benchmark 提出MMLU-CF：一个无污染的多任务语言理解基准，用于更可靠地评估LLM。 large language model ✅
24 Reasoning Through Execution: Unifying Process and Outcome Rewards for Code Generation 提出Outcome Refining Process Supervision，统一过程和结果奖励，提升代码生成质量。 large language model ✅
25 ALKAFI-LLAMA3: Fine-Tuning LLMs for Precise Legal Understanding in Palestine ALKAFI-LLAMA3：微调LLM以实现巴勒斯坦法律的精准理解 large language model
26 Language Models as Continuous Self-Evolving Data Engineers 提出LANCE：一种基于LLM的持续自进化数据工程框架，提升模型性能。 large language model ✅
27 Dehallucinating Parallel Context Extension for Retrieval-Augmented Generation 提出DePaC，通过解耦幻觉缓解RAG中并行上下文扩展的问题 large language model
28 How good is GPT at writing political speeches for the White House? 评估GPT在撰写白宫政治演讲稿方面的能力：对比GPT与美国总统的演讲风格 large language model
29 All-in-One Tuning and Structural Pruning for Domain-Specific LLMs 提出ATP：面向领域LLM的端到端调优与结构化剪枝方法 large language model
30 SKETCH: Structured Knowledge Enhanced Text Comprehension for Holistic Retrieval SKETCH：融合结构化知识的文本理解方法，提升RAG系统检索性能 large language model
31 Automatic Extraction of Metaphoric Analogies from Literary Texts: Task Formulation, Dataset Construction, and Evaluation 提出文学文本隐喻类比自动抽取方法，并构建数据集用于评估大型语言模型。 large language model
32 Decade of Natural Language Processing in Chronic Pain: A Systematic Review 综述：自然语言处理在慢性疼痛研究中的十年进展与未来方向 multimodal
33 ConfliBERT: A Language Model for Political Conflict ConfliBERT：用于政治冲突事件抽取的专用语言模型，性能超越通用LLM。 large language model
34 LLMs Lost in Translation: M-ALERT uncovers Cross-Linguistic Safety Inconsistencies M-ALERT揭示LLM多语言安全漏洞，发现跨语言安全一致性问题 large language model
35 Chain-of-MetaWriting: Linguistic and Textual Analysis of How Small Language Models Write Young Students Texts 提出Chain-of-MetaWriting方法，分析小型语言模型在辅助青少年写作中的表现与局限 large language model
36 Think&Cite: Improving Attributed Text Generation with Self-Guided Tree Search and Progress Reward Modeling 提出Think&Cite框架，通过自引导树搜索和进度奖励建模提升属性文本生成的事实准确性。 large language model
37 ViFactCheck: A New Benchmark Dataset and Methods for Multi-domain News Fact-Checking in Vietnamese ViFactCheck：提出越南语多领域新闻事实核查基准数据集与方法 large language model
38 Disentangling Reasoning Tokens and Boilerplate Tokens For Language Model Fine-tuning 提出RFT方法，通过解耦推理和样板 tokens，提升LLM在agent任务上的微调效果。 large language model
39 On Verbalized Confidence Scores for LLMs 提出一种提示工程方法，使LLM能够输出校准良好的置信度评分，用于不确定性量化。 large language model ✅
40 Beyond Guilt: Legal Judgment Prediction with Trichotomous Reasoning 提出LJPIV基准数据集，增强法律LLM的三分式推理能力，提升无罪判决预测准确性。 large language model
41 Agent-SafetyBench: Evaluating the Safety of LLM Agents Agent-SafetyBench：构建LLM Agent安全评估基准，揭示现有Agent安全风险 large language model ✅
42 To Err Is Human; To Annotate, SILICON? Reducing Measurement Error in LLM Annotation 提出SILICON方法，系统性降低LLM文本标注中的测量误差，提升管理研究的标注质量和可复现性。 large language model

🔬 支柱二：RL算法与架构 (RL & Architecture) (10 篇)

#题目一句话要点标签🔗⭐
43 Qwen2.5 Technical Report Qwen2.5：通过扩展数据和优化训练，显著提升大语言模型性能 reinforcement learning large language model multimodal
44 Do Large Language Models Advocate for Inferentialism? 探讨大型语言模型是否支持推论主义语义学理论 RLHF large language model
45 Northeastern Uni at Multilingual Counterspeech Generation: Enhancing Counter Speech Generation with LLM Alignment through Direct Preference Optimization 利用直接偏好优化对齐LLM，提升多语种反制言论生成效果 DPO direct preference optimization large language model
46 Efficient Knowledge Injection in LLMs via Self-Distillation 提出基于自蒸馏的prompt distillation方法，高效地将新知识注入大语言模型 distillation large language model
47 Self-Evolution Knowledge Distillation for LLM-based Machine Translation 提出自进化知识蒸馏方法，提升基于LLM的机器翻译性能 distillation large language model
48 CORD: Balancing COnsistency and Rank Distillation for Robust Retrieval-Augmented Generation CORD：平衡一致性与排序蒸馏，提升检索增强生成模型的鲁棒性 distillation large language model
49 Multi-Level Optimal Transport for Universal Cross-Tokenizer Knowledge Distillation on Language Models 提出多层最优传输方法，实现语言模型跨分词器的知识蒸馏。 distillation large language model ✅
50 PA-RAG: RAG Alignment via Multi-Perspective Preference Optimization 提出PA-RAG，通过多视角偏好优化对齐RAG生成器，提升信息性、鲁棒性和引用质量。 DPO direct preference optimization large language model ✅
51 LDC: Learning to Generate Research Idea with Dynamic Control 提出LDC框架，通过动态控制生成高质量科研想法，平衡新颖性、可行性和有效性。 reinforcement learning large language model
52 Simulation-Free Hierarchical Latent Policy Planning for Proactive Dialogues 提出LDPP，一种无仿真分层潜在策略规划框架，用于主动对话。 reinforcement learning large language model

🔬 支柱一：机器人控制 (Robot Control) (1 篇)

#题目一句话要点标签🔗⭐
53 Mapping and Influencing the Political Ideology of Large Language Models using Synthetic Personas 利用合成角色探究并操控大语言模型的政治倾向 manipulation large language model

⬅️ 返回 cs.CL 首页 · 🏠 返回主页