cs.CL（2025-03-04）

📊 共 45 篇论文 | 🔗 13 篇有代码

🎯 兴趣领域导航

支柱九：具身大模型 (Embodied Foundation Models) (37 🔗9) 支柱二：RL算法与架构 (RL & Architecture) (8 🔗4)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (37 篇)

#题目一句话要点标签🔗⭐
1 InSerter: Speech Instruction Following with Unsupervised Interleaved Pre-training 提出InSerter：一种基于非监督交错预训练的语音指令跟随方法 large language model instruction following
2 MCiteBench: A Multimodal Benchmark for Generating Text with Citations 提出MCiteBench，评估多模态大语言模型生成带引用文本的能力，解决幻觉问题。 large language model multimodal
3 InfiniSST: Simultaneous Translation of Unbounded Speech with Large Language Model InfiniSST：利用大语言模型实现无界语音的同步翻译 large language model ✅
4 Multilingual Relative Clause Attachment Ambiguity Resolution in Large Language Models 研究大型语言模型在多语种环境下关系从句附着歧义消解能力，揭示其跨语言处理差异。 large language model
5 Large Language Models for Multilingual Previously Fact-Checked Claim Detection 评估大型语言模型在多语种已核实声明检测中的性能 large language model
6 Generator-Assistant Stepwise Rollback Framework for Large Language Model Agent 提出GA-Rollback框架，解决LLM Agent一步到位推理中的错误传播问题 large language model
7 AILS-NTUA at SemEval-2025 Task 4: Parameter-Efficient Unlearning for Large Language Models using Data Chunking 提出基于数据分块的参数高效LLM遗忘方法，解决敏感内容移除问题。 large language model
8 AILS-NTUA at SemEval-2025 Task 3: Leveraging Large Language Models and Translation Strategies for Multilingual Hallucination Detection 利用大语言模型和翻译策略解决多语言幻觉检测问题 large language model
9 MedEthicEval: Evaluating Large Language Models Based on Chinese Medical Ethics MedEthicEval：构建中文医学伦理评估基准，评估大型语言模型伦理推理能力 large language model
10 Add-One-In: Incremental Sample Selection for Large Language Models via a Choice-Based Greedy Paradigm 提出Add-One-In增量采样法，利用LLM进行高质量、多样性大模型训练数据选择。 large language model ✅
11 PromptCoT: Synthesizing Olympiad-level Problems for Mathematical Reasoning in Large Language Models PromptCoT：合成奥林匹克级别数学题，提升大语言模型数学推理能力 large language model ✅
12 FairSense-AI: Responsible AI Meets Sustainability FairSense-AI：一个负责任且可持续的多模态AI偏见检测与缓解框架 large language model multimodal ✅
13 OmniSQL: Synthesizing High-quality Text-to-SQL Data at Scale 提出OmniSQL框架，大规模合成高质量Text-to-SQL数据，并训练开源模型。 large language model chain-of-thought
14 Prompting Science Report 1: Prompt Engineering is Complicated and Contingent 提示工程复杂且依赖情境：基准测试标准选择和提示策略对大语言模型性能影响显著 large language model
15 Multi-Agent System for AI-Assisted Extraction of Narrative Arcs in TV Series 提出一种多智能体系统，用于AI辅助提取电视剧叙事弧线。 multimodal
16 LINGOLY-TOO: Disentangling Reasoning from Knowledge with Templatised Orthographic Obfuscation 提出LINGOLY-TOO基准，通过模板化正字法混淆解耦语言模型中的推理与知识 large language model
17 Implicit Bias in LLMs: A Survey 综述隐性偏见在大型语言模型中的影响及检测方法 large language model
18 MPO: Boosting LLM Agents with Meta Plan Optimization MPO：通过元计划优化提升LLM Agent能力 large language model
19 Improving LLM-as-a-Judge Inference with the Judgment Distribution 利用判断分布改进LLM作为裁判的推理性能 chain-of-thought
20 SAFE: A Sparse Autoencoder-Based Framework for Robust Query Enrichment and Hallucination Mitigation in LLMs 提出SAFE框架，利用稀疏自编码器增强LLM查询并缓解幻觉问题 large language model
21 The First Few Tokens Are All You Need: An Efficient and Effective Unsupervised Prefix Fine-Tuning Method for Reasoning Models 提出无监督前缀微调(UPFT)，高效提升LLM推理能力，无需标注数据。 large language model
22 SteerConf: Steering LLMs for Confidence Elicitation SteerConf：通过引导LLM置信度来提高校准性和可靠性 large language model
23 Shakespearean Sparks: The Dance of Hallucination and Creativity in LLMs' Decoding Layers 提出HCL框架，量化评估LLM解码层中的幻觉与创造力权衡 large language model ✅
24 Multidimensional Consistency Improves Reasoning in Language Models 多维度一致性提升语言模型推理能力，尤其对数学问题 large language model
25 Put the Space of LoRA Initialization to the Extreme to Preserve Pre-trained Knowledge 提出LoRA-Null，通过激活空空间初始化LoRA，有效缓解大语言模型微调中的灾难性遗忘。 large language model ✅
26 Towards Event Extraction with Massive Types: LLM-based Collaborative Annotation and Partitioning Extraction 提出基于LLM的协同标注与划分抽取方法，用于解决大规模类型事件抽取问题。 large language model
27 LADM: Long-context Training Data Selection with Attention-based Dependency Measurement for LLMs 提出LADM框架，利用注意力机制进行长文本数据选择，提升LLM长文本处理能力。 large language model
28 Measuring What Makes You Unique: Difference-Aware User Modeling for Enhancing LLM Personalization 提出差异感知个性化学习(DPL)，增强LLM的个性化能力 large language model ✅
29 Hierarchical Re-ranker Retriever (HRR) 提出分层重排序检索器(HRR)，解决LLM应用中上下文检索粒度选择难题 large language model
30 An Efficient and Precise Training Data Construction Framework for Process-supervised Reward Model in Mathematical Reasoning EpicPRM：高效精确的数学推理过程监督奖励模型训练数据构建框架 large language model ✅
31 PanguIR Technical Report for NTCIR-18 AEOLLM Task PanguIR提出多模型协作、提示自动优化和ICL优化方法，提升LLM自动评估性能。 large language model
32 DeLTa: A Decoding Strategy based on Logit Trajectory Prediction Improves Factuality and Reasoning Ability DeLTa：基于Logit轨迹预测的解码策略，提升大语言模型的真实性和推理能力 large language model
33 Limited Effectiveness of LLM-based Data Augmentation for COVID-19 Misinformation Stance Detection 评估LLM数据增强在COVID-19虚假信息立场检测中的有效性，发现其增益有限 large language model
34 Haste Makes Waste: Evaluating Planning Abilities of LLMs for Efficient and Feasible Multitasking with Time Constraints Between Actions 提出Recipe2Plan基准，评估LLM在时序约束下高效多任务规划能力 large language model ✅
35 Enhancing LLM Reliability via Explicit Knowledge Boundary Modeling 提出显式知识边界建模框架以提升大型语言模型的可靠性 large language model
36 Call for Rigor in Reporting Quality of Instruction Tuning Data 强调Instruction Tuning数据质量评估中超参数选择严谨性的必要性 large language model
37 Measuring Intrinsic Dimension of Token Embeddings 通过测量token嵌入的本征维度评估语言模型的冗余度并指导LoRA应用 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🔗⭐
38 Rewarding Doubt: A Reinforcement Learning Approach to Calibrated Confidence Expression of Large Language Models 提出基于强化学习的LLM置信度校准方法，提升模型回答事实性问题的可靠性。 reinforcement learning reward design large language model
39 AlignDistil: Token-Level Language Model Alignment as Adaptive Policy Distillation AlignDistil：提出一种基于Token级别语言模型对齐的自适应策略蒸馏方法 reinforcement learning RLHF DPO
40 Mask-DPO: Generalizable Fine-grained Factuality Alignment of LLMs 提出Mask-DPO，通过细粒度事实性对齐提升大语言模型的事实准确性 preference learning DPO direct preference optimization
41 Learning from Failures in Multi-Attempt Reinforcement Learning 提出多尝试强化学习，提升LLM在复杂推理任务中的表现 reinforcement learning large language model ✅
42 ATLaS: Agent Tuning via Learning Critical Steps ATLaS：通过学习关键步骤进行Agent调优，提升LLM Agent泛化能力 behavior cloning generalist agent large language model
43 SAGE: Steering Dialog Generation with Future-Aware State-Action Augmentation SAGE：利用未来感知的状态-动作增强来引导对话生成 reinforcement learning SAC large language model ✅
44 It Helps to Take a Second Opinion: Teaching Smaller LLMs to Deliberate Mutually via Selective Rationale Optimisation 提出COALITION框架，通过选择性推理优化提升小型LLM在复杂任务中的推理能力。 distillation large language model ✅
45 Teaching Your Models to Understand Code via Focal Preference Alignment 提出Target-DPO，通过焦点偏好对齐提升代码大模型理解代码能力 preference learning DPO ✅

⬅️ 返回 cs.CL 首页 · 🏠 返回主页