cs.CL（2026-02-05）

📊 共 35 篇论文 | 🔗 7 篇有代码

🎯 兴趣领域导航

支柱九：具身大模型 (Embodied Foundation Models) (27 🔗6) 支柱二：RL算法与架构 (RL & Architecture) (7 🔗1) 支柱一：机器人控制 (Robot Control) (1)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (27 篇)

#题目一句话要点标签🔗⭐
1 Cross-Lingual Empirical Evaluation of Large Language Models for Arabic Medical Tasks 跨语言评估大型语言模型在阿拉伯语医疗任务中的表现差距 large language model
2 Transport and Merge: Cross-Architecture Merging for Large Language Models 提出基于最优传输的跨架构模型融合方法，实现大模型知识向异构小模型的迁移。 large language model
3 A Systematic Evaluation of Large Language Models for PTSD Severity Estimation: The Role of Contextual Knowledge and Modeling Strategies 系统评估大型语言模型在创伤后应激障碍严重程度评估中的作用，着重上下文知识和建模策略的影响。 large language model
4 OdysseyArena: Benchmarking Large Language Models For Long-Horizon, Active and Inductive Interactions OdysseyArena：用于长时程、主动和归纳交互的大语言模型基准测试 large language model ✅
5 Consensus-Aligned Neuron Efficient Fine-Tuning Large Language Models for Multi-Domain Machine Translation 提出共识对齐神经元高效微调方法，提升大语言模型在多领域机器翻译中的性能 large language model
6 CASTLE: A Comprehensive Benchmark for Evaluating Student-Tailored Personalized Safety in Large Language Models 提出CASTLE：一个综合性评测基准，用于评估大语言模型中面向学生的个性化安全。 large language model
7 SciDef: Automating Definition Extraction from Academic Literature with Large Language Models SciDef：提出一种基于大语言模型的学术文献定义自动抽取流程。 large language model ✅
8 OPUS: Towards Efficient and Principled Data Selection in Large Language Model Pre-training in Every Iteration OPUS：通过优化器引导的投影效用选择，实现大语言模型预训练的迭代式高效数据选择。 large language model
9 IESR:Efficient MCTS-Based Modular Reasoning for Text-to-SQL with Large Language Models IESR：一种高效的基于MCTS的模块化推理框架，用于大型语言模型上的Text-to-SQL任务 large language model ✅
10 EuroLLM-22B: Technical Report EuroLLM-22B：为欧洲公民需求从头训练的多语言大语言模型 large language model instruction following
11 RRAttention: Dynamic Block Sparse Attention via Per-Head Round-Robin Shifts for Long-Context Inference RRAttention：通过轮询移位实现动态块稀疏注意力，加速长文本推理。 large language model multimodal
12 Multi-Task GRPO: Reliable LLM Reasoning Across Tasks 提出MT-GRPO算法，提升LLM在多任务场景下的可靠推理性能，尤其关注最差任务表现。 large language model
13 Late-to-Early Training: LET LLMs Learn Earlier, So Faster and Better 提出LET：利用小模型知识加速大语言模型训练，提升性能。 large language model
14 CoPE: Clipped RoPE as A Scalable Free Lunch for Long Context LLMs 提出CoPE：一种可扩展的RoPE软裁剪方法，提升长文本LLM性能 large language model ✅
15 Are Open-Weight LLMs Ready for Social Media Moderation? A Comparative Study on Bluesky 评估开源LLM在社交媒体审核中的应用：以Bluesky平台为例 large language model
16 LinguistAgent: A Reflective Multi-Model Platform for Automated Linguistic Annotation LinguistAgent：一个用于自动语言学标注的反射式多模型平台 large language model ✅
17 Structured Context Engineering for File-Native Agentic Systems: Evaluating Schema Accuracy, Format Effectiveness, and Multi-File Navigation at Scale 针对文件原生Agent系统，研究结构化上下文工程对SQL生成任务的影响 large language model
18 Towards a Science of Collective AI: LLM-based Multi-Agent Systems Need a Transition from Blind Trial-and-Error to Rigorous Science 构建集体AI科学体系：提出基于LLM的多智能体系统从盲目试错到严谨科学的过渡框架 large language model
19 DFlash: Block Diffusion for Flash Speculative Decoding DFlash：提出基于块扩散的Flash推测解码框架，加速LLM推理。 large language model
20 DSB: Dynamic Sliding Block Scheduling for Diffusion LLMs 提出动态滑动块调度（DSB）方法，提升Diffusion LLM的生成质量和效率。 large language model ✅
21 KV-CoRE: Benchmarking Data-Dependent Low-Rank Compressibility of KV-Caches in LLMs KV-CoRE：一种评估LLM中KV-Cache数据依赖性低秩可压缩性的基准方法 large language model
22 Codified Finite-state Machines for Role-playing 提出Codified有限状态机，提升LLM角色扮演的连贯性和可控性 large language model
23 Causal Front-Door Adjustment for Robust Jailbreak Attacks on LLMs 提出CFA²框架，利用因果前门调整实现对LLM的鲁棒越狱攻击 large language model
24 Once Correct, Still Wrong: Counterfactual Hallucination in Multilingual Vision-Language Models 提出M2CQA基准与CFHR指标，揭示多语言视觉-语言模型在文化背景下的反事实幻觉问题 multimodal
25 Multi-Field Tool Retrieval 提出多字段工具检索框架，解决LLM工具检索中的语义鸿沟与多维度建模问题。 large language model
26 FedMosaic: Federated Retrieval-Augmented Generation via Parametric Adapters FedMosaic：基于参数化适配器的联邦检索增强生成框架，解决隐私场景下的知识孤岛问题。 large language model
27 Bagpiper: Solving Open-Ended Audio Tasks via Rich Captions Bagpiper：通过富文本描述解决开放域音频任务的80亿参数音频基础模型 foundation model

🔬 支柱二：RL算法与架构 (RL & Architecture) (7 篇)

#题目一句话要点标签🔗⭐
28 Aligning Large Language Model Behavior with Human Citation Preferences 对齐大语言模型行为与人类引文偏好，提升LLM生成内容的可信度 direct preference optimization large language model
29 Stop Rewarding Hallucinated Steps: Faithfulness-Aware Step-Level Reinforcement Learning for Small Reasoning Models 提出FaithRL，解决小模型推理中中间步骤的幻觉问题，提升推理可靠性 reinforcement learning large language model chain-of-thought ✅
30 Length-Unbiased Sequence Policy Optimization: Revealing and Controlling Response Length Variation in RLVR 提出长度无偏序列策略优化(LUSPO)，解决RLVR中响应长度崩溃问题。 reinforcement learning large language model multimodal
31 PACE: Defying the Scaling Hypothesis of Exploration in Iterative Alignment for Mathematical Reasoning PACE：数学推理迭代对齐中，挑战探索规模假设，提升效率与鲁棒性 DPO direct preference optimization large language model
32 Learning Query-Aware Budget-Tier Routing for Runtime Agent Memory 提出BudgetMem，通过查询感知的预算分层路由优化LLM Agent运行时记忆。 reinforcement learning large language model
33 Multi-Token Prediction via Self-Distillation 提出基于自蒸馏的多Token预测方法，加速语言模型推理。 distillation
34 LongR: Unleashing Long-Context Reasoning via Reinforcement Learning with Dense Utility Rewards LongR：利用强化学习与密集效用奖励，释放长文本推理能力 reinforcement learning

🔬 支柱一：机器人控制 (Robot Control) (1 篇)

#题目一句话要点标签🔗⭐
35 Reinforcement World Model Learning for LLM-based Agents 提出RWML，通过强化世界模型学习提升LLM智能体在文本环境中的决策能力。 sim-to-real world model large language model

⬅️ 返回 cs.CL 首页 · 🏠 返回主页