cs.AI（2026-02-12）

📊 共 43 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) (29 🔗5) 支柱二：RL算法与架构 (RL & Architecture) (13 🔗2) 支柱一：机器人控制 (Robot Control) (1)

🔬 支柱九：具身大模型 (Embodied Foundation Models) (29 篇)

#题目一句话要点标签🔗⭐
1 Commencing-Student Enrolment Forecasting Under Data Sparsity with Time Series Foundation Models 利用时间序列基础模型，解决数据稀疏下的高校新生入学预测问题 foundation model
2 InjectRBP: Steering Large Language Model Reasoning Behavior via Pattern Injection InjectRBP：通过行为模式注入引导大语言模型推理 large language model
3 Talk2DM: Enabling Natural Language Querying and Commonsense Reasoning for Vehicle-Road-Cloud Integrated Dynamic Maps with Large Language Models 提出Talk2DM，通过自然语言查询和常识推理增强车-路-云协同动态地图 large language model
4 scPilot: Large Language Model Reasoning Toward Automated Single-Cell Analysis and Discovery scPilot：利用大语言模型推理实现自动化单细胞分析与发现 large language model ✅
5 Differentially Private and Communication Efficient Large Language Model Split Inference via Stochastic Quantization and Soft Prompt 提出DEL框架，通过差分隐私随机量化和软提示实现通信高效的LLM分割推理。 large language model
6 Beyond Pixels: Vector-to-Graph Transformation for Reliable Schematic Auditing 提出向量-图转换以解决工程图纸结构盲目性问题 large language model multimodal ✅
7 Beyond Parameter Arithmetic: Sparse Complementary Fusion for Distribution-Aware Model Merging 提出SCF-RKL，通过稀疏互补融合解决模型合并中的功能干扰问题。 large language model instruction following
8 Do MLLMs Really Understand Space? A Mathematical Reasoning Evaluation MathSpatial：用于评估和提升多模态大语言模型空间数学推理能力的统一框架 large language model multimodal
9 Think like a Scientist: Physics-guided LLM Agent for Equation Discovery KeplerAgent：基于物理先验知识的LLM智能体，用于符号方程发现 large language model
10 GPT-4o Lacks Core Features of Theory of Mind GPT-4o缺乏核心的心智理论能力，无法建立连贯一致的心理状态模型 large language model
11 AttentionRetriever: Attention Layers are Secretly Long Document Retrievers 提出AttentionRetriever，利用注意力机制进行高效长文档检索。 large language model
12 VIRENA: Virtual Arena for Research, Education, and Democratic Innovation VIRENA：用于研究、教育和民主创新的虚拟社交媒体实验平台 large language model
13 Sci-CoE: Co-evolving Scientific Reasoning LLMs via Geometric Consensus with Sparse Supervision 提出Sci-CoE框架，通过几何共识与稀疏监督协同进化科学推理LLM large language model ✅
14 The Pensieve Paradigm: Stateful Language Models Mastering Their Own Context 提出StateLM，赋予语言模型记忆管理能力，提升长文本处理和对话性能。 foundation model
15 ModelWisdom: An Integrated Toolkit for TLA+ Model Visualization, Digest and Repair ModelWisdom：集成TLA+模型可视化、理解与修复工具，提升模型检查效率。 large language model
16 IncompeBench: A Permissively Licensed, Fine-Grained Benchmark for Music Information Retrieval IncompeBench：一个许可宽松、细粒度的音乐信息检索评测基准。 multimodal ✅
17 Leveraging LLMs to support co-evolution between definitions and instances of textual DSLs: A Systematic Evaluation 利用LLM协同演化文本DSL定义与实例，系统评估其性能与局限性 large language model
18 From Atoms to Trees: Building a Structured Feature Forest with Hierarchical Sparse Autoencoders 提出HSAE，通过分层稀疏自编码器构建结构化特征森林，挖掘LLM中的多尺度概念结构。 large language model
19 AmbiBench: Benchmarking Mobile GUI Agents Beyond One-Shot Instructions in the Wild AmbiBench：构建移动GUI Agent基准，评估其在真实场景下处理模糊指令和意图对齐的能力。 instruction following
20 AIR: Improving Agent Safety through Incident Response AIR：通过事件响应提升LLM Agent的安全性 large language model
21 Text2GQL-Bench: A Text to Graph Query Language Benchmark [Experiment, Analysis & Benchmark] 提出Text2GQL-Bench，用于评估和提升文本到图查询语言的转换性能 large language model
22 Benchmark Health Index: A Systematic Framework for Benchmarking the Benchmarks of LLMs 提出基准健康指数BHI，用于系统性评估和管理LLM基准的可靠性。 large language model
23 PhyNiKCE: A Neurosymbolic Agentic Framework for Autonomous Computational Fluid Dynamics PhyNiKCE：一种神经符号代理框架，用于自主计算流体动力学。 large language model
24 LoRA-based Parameter-Efficient LLMs for Continuous Learning in Edge-based Malware Detection 提出基于LoRA的参数高效LLM持续学习框架，用于边缘恶意软件检测。 large language model
25 MAPLE: Modality-Aware Post-training and Learning Ecosystem 提出MAPLE，通过模态感知后训练提升多模态强化学习性能。 multimodal
26 SemaPop: Semantic-Persona Conditioned Population Synthesis SemaPop：提出一种语义-角色条件的人口合成方法，融合LLM与生成模型。 large language model
27 Stop Tracking Me! Proactive Defense Against Attribute Inference Attack in LLMs 提出TRACE-RPS框架，主动防御LLM中的属性推断攻击 large language model ✅
28 AgentLeak: A Full-Stack Benchmark for Privacy Leakage in Multi-Agent LLM Systems AgentLeak：多智能体LLM系统隐私泄露的全栈基准测试 large language model
29 Compiler-Guided Inference-Time Adaptation: Improving GPT-5 Programming Performance in Idris 编译器指导的推理时自适应：提升GPT-5在Idris编程中的性能 large language model

🔬 支柱二：RL算法与架构 (RL & Architecture) (13 篇)

#题目一句话要点标签🔗⭐
30 Budget-Constrained Agentic Large Language Models: Intention-Based Planning for Costly Tool Use 提出INTENT框架以解决预算约束下的工具使用问题 world model large language model
31 Provable Offline Reinforcement Learning for Structured Cyclic MDPs 提出CycleFQI，解决结构化循环MDPs的离线强化学习问题 reinforcement learning offline reinforcement learning
32 Pedagogically-Inspired Data Synthesis for Language Model Knowledge Distillation 提出IOA框架，通过教学启发的数据合成方法提升语言模型知识蒸馏效果 distillation large language model
33 TSR: Trajectory-Search Rollouts for Multi-Turn RL of LLM Agents TSR：用于LLM Agent多轮RL的轨迹搜索Rollout方法 reinforcement learning PPO large language model
34 Tiny Recursive Reasoning with Mamba-2 Attention Hybrid Mamba-2混合注意力机制提升了Tiny递归推理模型在抽象推理任务上的性能 Mamba SSM
35 Predicting LLM Output Length via Entropy-Guided Representations 提出基于熵引导表征的LLM输出长度预测框架，提升推理效率。 reinforcement learning MAE chain-of-thought
36 PuYun-LDM: A Latent Diffusion Model for High-Resolution Ensemble Weather Forecasts PuYun-LDM：一种用于高分辨率集合天气预报的潜在扩散模型 masked autoencoder MAE foundation model
37 RELATE: A Reinforcement Learning-Enhanced LLM Framework for Advertising Text Generation 提出RELATE框架，利用强化学习增强LLM，解决广告文本生成中目标不一致问题。 reinforcement learning policy learning
38 The Five Ws of Multi-Agent Communication: Who Talks to Whom, When, What, and Why -- A Survey from MARL to Emergent Language and LLMs 多智能体通信五要素综述：从MARL到涌现语言和LLM reinforcement learning large language model
39 CM2: Reinforcement Learning with Checklist Rewards for Multi-Turn and Multi-Step Agentic Tool Use CM2：基于清单奖励的强化学习，用于多轮多步Agentic工具使用 reinforcement learning ✅
40 SAM3-LiteText: An Anatomical Study of the SAM3 Text Encoder for Efficient Vision-Language Segmentation 提出SAM3-LiteText，通过知识蒸馏压缩文本编码器，提升视觉-语言分割效率。 distillation visual grounding ✅
41 Seq2Seq2Seq: Lossless Data Compression via Discrete Latent Transformers and Reinforcement Learning 提出基于离散潜在Transformer和强化学习的Seq2Seq2Seq无损数据压缩方法 reinforcement learning
42 HLA: Hadamard Linear Attention 提出Hadamard线性注意力(HLA)，用于高效近似Transformer中的softmax注意力机制。 linear attention

🔬 支柱一：机器人控制 (Robot Control) (1 篇)

#题目一句话要点标签🔗⭐
43 Gaia2: Benchmarking LLM Agents on Dynamic and Asynchronous Environments Gaia2：用于评估LLM Agent在动态异步环境中性能的基准测试 sim2real reinforcement learning large language model

⬅️ 返回 cs.AI 首页 · 🏠 返回主页