cs.CL（2026-05-13）

📊 共 32 篇论文 | 🔗 4 篇有代码

🎯 兴趣领域导航

支柱九：具身大模型 (Embodied Foundation Models) (26 🔗4) 支柱二：RL算法与架构 (RL & Architecture) (6)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (26 篇)

#题目一句话要点标签🔗⭐
1 DiM\textsuperscript{3}: Bridging Multilingual and Multimodal Models via Direction- and Magnitude-Aware Merging 提出DiM³，通过方向和幅度感知的融合方法，桥接多语言和多模态模型。 large language model multimodal ✅
2 Utility-Oriented Visual Evidence Selection for Multimodal Retrieval-Augmented Generation 提出面向效用的视觉证据选择方法，提升多模态检索增强生成效果 multimodal
3 Continual Learning with Multilingual Foundation Model 提出多阶段框架，利用多语言基础模型检测社交媒体中LGBTQ+相关侮辱性词语的赎回用法。 foundation model ✅
4 Query-Conditioned Test-Time Self-Training for Large Language Models 提出查询条件自训练QueST，提升大语言模型在推理任务中的测试时自适应能力 large language model
5 FIND: Toward Multimodal Financial Reasoning and Question Answering for Indic Languages 提出FIND框架，解决印地语多模态金融推理与问答难题。 multimodal
6 RAG-Enhanced Large Language Models for Dynamic Content Expiration Prediction in Web Search 提出基于RAG增强的大语言模型，用于Web搜索中动态内容过期预测 large language model
7 Leveraging Multimodal Self-Consistency Reasoning in Coding Motivational Interviewing for Alcohol Use Reduction 提出基于多模态自洽性推理的酒精使用干预自动编码方法，提升编码鲁棒性。 multimodal
8 CommonWhy: A Dataset for Evaluating Entity-Based Causal Commonsense Reasoning in Large Language Models 提出CommonWhy数据集，用于评估大语言模型中基于实体的因果常识推理能力 large language model
9 CiteVQA: Benchmarking Evidence Attribution for Trustworthy Document Intelligence 提出CiteVQA基准，用于评估文档智能中证据溯源的可信度，揭示现有模型中的“溯源幻觉”问题。 large language model multimodal ✅
10 Many-Shot CoT-ICL: Making In-Context Learning Truly Learn 提出Curvilinear Demonstration Selection (CDS)方法，提升CoT-ICL在推理任务上的性能。 large language model chain-of-thought
11 GeoBuildBench: A Benchmark for Interactive and Executable Geometry Construction from Natural Language 提出GeoBuildBench：一个用于交互式几何构造的自然语言基准测试。 large language model multimodal
12 Children's English Reading Story Generation via Supervised Fine-Tuning of Compact LLMs with Controllable Difficulty and Safety 通过监督式微调小型LLM，实现可控难度和安全性的儿童英语阅读故事生成。 large language model
13 LLMs as Implicit Imputers: Uncertainty Should Scale with Missing Information 将LLM视为隐式插补器，提出不确定性应随缺失信息量增加的评估准则。 large language model
14 WARDEN: Endangered Indigenous Language Transcription and Translation with 6 Hours of Training Data 提出WARDEN以解决濒危土著语言转录与翻译问题 large language model
15 LLMs as annotators of credibility assessment in Danish asylum decisions: evaluating classification performance and errors beyond aggregated metrics 利用LLM标注丹麦语庇护决策可信度评估：评估分类性能与误差 large language model ✅
16 Probing Persona-Dependent Preferences in Language Models 通过线性探针揭示语言模型中人格依赖的偏好机制 large language model
17 Tracing Persona Vectors Through LLM Pretraining 追踪LLM预训练中的Persona向量：揭示早期形成与持续优化机制 large language model
18 Context Training with Active Information Seeking 提出基于主动信息搜索的上下文训练方法，提升LLM在特定领域任务上的性能 large language model
19 Persona-Model Collapse in Emergent Misalignment 揭示涌现性错位中的人格模型崩溃现象，并提出敏感诊断指标 large language model
20 An LLM-Based System for Argument Reconstruction 提出基于LLM的端到端系统，用于从自然语言文本中重构论证结构。 large language model
21 Edit-level Majority Voting Mitigates Over-Correction in LLM-based Grammatical Error Correction 提出基于编辑级别多数投票的推理方法，缓解LLM语法纠错中的过度纠正问题 large language model
22 Inducing Artificial Uncertainty in Language Models 提出一种人工诱导不确定性的方法，提升语言模型在安全关键应用中的不确定性量化能力。 large language model
23 PersonalAI 2.0: Enhancing knowledge graph traversal/retrieval with planning mechanism for Personalized LLM Agents PersonalAI 2.0：通过规划机制增强知识图谱遍历/检索，用于个性化LLM Agent large language model
24 Model-Agnostic Lifelong LLM Safety via Externalized Attack-Defense Co-Evolution EvoSafety：通过外部化攻防协同进化实现模型无关的终身LLM安全 large language model
25 From Rosetta to Match-Up: A Paired Corpus of Linguistic Puzzles with Human and LLM Benchmarks 提出系统化方法将Rosetta Stone难题转换为Match-Up格式 large language model
26 The Cost of Perfect English: Pragmatic Flattening and the Erasure of Authorial Voice in L2 Writing Supported by GenAI 研究揭示GenAI润色L2写作中存在的语用扁平化及作者声音的消解问题 large language model

🔬 支柱二：RL算法与架构 (RL & Architecture) (6 篇)

#题目一句话要点标签🔗⭐
27 GateKD: Confidence-Gated Closed-Loop Distillation for Robust Reasoning 提出GateKD：一种置信度门控的闭环蒸馏框架，用于提升鲁棒推理能力。 teacher-student distillation large language model
28 Prefix Teach, Suffix Fade: Local Teachability Collapse in Strong-to-Weak On-Policy Distillation 提出局部可教性崩塌现象，并设计自适应截断策略优化强弱On-Policy蒸馏。 teacher-student distillation
29 GAGPO: Generalized Advantage Grouped Policy Optimization 提出GAGPO，一种免Critic的强化学习方法，解决多轮交互环境中的信用分配难题。 reinforcement learning large language model
30 Where Does Reasoning Break? Step-Level Hallucination Detection via Hidden-State Transport Geometry 提出基于隐状态轨迹几何的步骤级幻觉检测方法，解决多步推理中的错误定位问题。 distillation large language model
31 TokAlign++: Advancing Vocabulary Adaptation via Better Token Alignment TokAlign++：通过改进Token对齐提升词汇适配性能 distillation large language model
32 STOP: Structured On-Policy Pruning of Long-Form Reasoning in Low-Data Regimes 提出STOP算法，用于低数据场景下长链推理的结构化策略剪枝。 distillation chain-of-thought

⬅️ 返回 cs.CL 首页 · 🏠 返回主页