ICML 2026将在2026年7月6日—11日于韩国首尔（Seoul, South Korea）举行。本文总结了2026 ICML上有关LLM × Graph相关论文。如有疏漏，欢迎大家补充。

注：笔者将分为上下2篇推文来总结，本文主要涉及针对图任务本身的的论文。

本文Graph的Topic：Graph4LLM，Graph4Agent，智能体记忆（Memory），AgenticRL，RAG等。

1. NaviAgent: Graph‑Driven Bilevel Planning for Scalable Tool Orchestration 2. Graph of States: Solving Abductive Tasks with Large Language Models 3. Beyond Trajectory-Level Attribution: Graph-Based Credit Assignment for Agentic Reinforcement Learning 4. Graph-R1: Towards Agentic GraphRAG Framework via End-to-end Reinforcement Learning 5. When Do Hallucinations Arise? A Graph Perspective on the Evolution of Path Reuse and Path Compression 6. MultiHal: Multilingual Dataset for Knowledge-Graph Grounded Evaluation of LLM Hallucinations 7. HugRAG: Hierarchical Causal Knowledge Graph Design for RAG 8. VimRAG: Navigating Massive Visual Context in Retrieval-Augmented Generation via Multimodal Memory Graph 9. From Retrieval to Translation: Translating Query into Graph-level Clues for Retrieval-Augmented Generation 10. Efficient Code Analysis via Graph-Guided Large Language Models 11. MASPOB: Bandit-Based Prompt Optimization for Multi-Agent Systems with Graph Neural Networks 12. Navigating the Energy Landscape of Collaboration: Multi-Agent Communication Graph Generation via Score-Based Diffusion 13. GraphFlow: A Graph-Based Workflow Management for Efficient LLM-Agent Serving 14. Factored Value Functions for Graph-Based Multi-Agent Reinforcement Learning 15. Embodied Task Planning via Graph-Informed Action Generation with Large Lanaguage Model 16. SAGE-NAS: Synergizing LLM-Based Semantic Agent with Graph-Based Evaluator for Neural Architecture Search 17. Weaving Graph over Tokens: Contextualizing Structured Sequences for LLMs 18. D${}^3$: Dynamic Directional Graph-Constrained Data Scheduling for LLM Training 19. GAUSS: Graph-Assisted Uncertainty Quantification using Structure and Semantics for Long-Form Generation in LLMs 20. Memory is Reconstructed, Not Retrieved: Graph Memory for LLM Agents

1 NaviAgent: Graph‑Driven Bilevel Planning for Scalable Tool Orchestration

链接：https://icml.cc/virtual/2026/poster/61578

arXiv：https://arxiv.org/abs/2506.19500

作者：Yan Jiang ⋅ HAO ZHOU ⋅ Lizhong Gu ⋅ Tianlong Li ⋅ Ruinan Jin ⋅ Wanqi Zhou ⋅ Ai Han

关键词：agent，工具编排，图驱动

2 Graph of States: Solving Abductive Tasks with Large Language Models

链接：https://icml.cc/virtual/2026/poster/65285

arXiv：https://arxiv.org/abs/2603.21250

代码：https://github.com/gaorch85/Graph-of-States

作者：Yu Luo ⋅ Rongchen Gao ⋅ Lu Teng ⋅ Xidao Wen ⋅ Jiamin Jiang ⋅ Qingliang Zhang ⋅ Yongqian Sun ⋅ Shenglin Zhang ⋅ Jiasong Feng ⋅ Tong Liu ⋅ Wenjie Zhang ⋅ Dan Pei

关键词：agent，逻辑推理

3 Beyond Trajectory-Level Attribution: Graph-Based Credit Assignment for Agentic Reinforcement Learning

链接：https://icml.cc/virtual/2026/poster/60543

作者：Xin Cheng ⋅ Shuo He ⋅ Lang Feng ⋅ Haiyang Xu ⋅ Ming Yan ⋅ Lei Feng ⋅ Bo An

关键词：GraphGPO，AgenticRL

4 Graph-R1: Towards Agentic GraphRAG Framework via End-to-end Reinforcement Learning

链接：https://icml.cc/virtual/2026/poster/63269

arXiv：http://arxiv.org/abs/2507.21892v1

作者：Haoran Luo ⋅ Haihong E ⋅ Guanting Chen ⋅ Qika Lin ⋅ Yikai Guo ⋅ Fangzhi Xu ⋅ Zemin Kuang ⋅ Meina Song ⋅ Xiaobao Wu ⋅ Yifan Zhu ⋅ Anh Tuan Luu

关键词：GraphRAG，AgenticRL，超图

5 When Do Hallucinations Arise? A Graph Perspective on the Evolution of Path Reuse and Path Compression

链接：https://icml.cc/virtual/2026/poster/66035

作者：Xinnan Dai ⋅ Kai Yang ⋅ cheng Luo ⋅ Shenglai Zeng ⋅ Kai Guo ⋅ Jiliang Tang

关键词：LLM幻觉，路径复用，路径压缩

6 MultiHal: Multilingual Dataset for Knowledge-Graph Grounded Evaluation of LLM Hallucinations

链接：https://icml.cc/virtual/2026/poster/62059

arXiv：https://arxiv.org/abs/2505.14101

作者：Ernests Lavrinovics ⋅ Russa Biswas ⋅ Katja Hose ⋅ Johannes Bjerva

关键词：LLM幻觉评估，多语言知识图谱

7 HugRAG: Hierarchical Causal Knowledge Graph Design for RAG

链接：https://icml.cc/virtual/2026/poster/65293

arXiv：https://arxiv.org/abs/2602.05143

作者：Nengbo Wang ⋅ Tuo Liang ⋅ Vikash Singh ⋅ Chaoda Song ⋅ Van Yang ⋅ Yu Yin ⋅ Jing Ma ⋅ JAGDIP SINGH ⋅ Vipin Chaudhary

关键词：RAG，多层级，知识图谱，因果

8 VimRAG: Navigating Massive Visual Context in Retrieval-Augmented Generation via Multimodal Memory Graph

链接：https://icml.cc/virtual/2026/poster/62507

arXiv：http://arxiv.org/abs/2602.12735v2

代码：https://github.com/Alibaba-NLP/VRAG

作者：Qiuchen Wang ⋅ Shihang Wang ⋅ Yu Zeng ⋅ Qiang Zhang ⋅ Fanrui Zhang ⋅ Zhuoning Guo ⋅ Bosi Zhang ⋅ Wenxuan Huang ⋅ Lin Chen ⋅ Zehui Chen ⋅ Pengjun Xie ⋅ Ruixue Ding

关键词：RAG，多模态图记忆

9 From Retrieval to Translation: Translating Query into Graph-level Clues for Retrieval-Augmented Generation

链接：https://icml.cc/virtual/2026/poster/65759

作者：Qichuan Liu ⋅ Chenfeng Zheng ⋅ Yuxuan Hu ⋅ Zerui Chen ⋅ Chentao Zhang ⋅ Qinggang Zhang ⋅ Zhihong Zhang

关键词：RAG，知识图谱

10 Efficient Code Analysis via Graph-Guided Large Language Models

链接：https://icml.cc/virtual/2026/poster/65144

arXiv：https://arxiv.org/abs/2601.12890

作者：Hang Gao ⋅ Tao Peng ⋅ Baoquan Cui ⋅ Hong Huang ⋅ Fengge Wu ⋅ Zhao Junsuo ⋅ Jian Zhang

关键词：代码分析，图引导的LLM

11 MASPOB: Bandit-Based Prompt Optimization for Multi-Agent Systems with Graph Neural Networks

链接：https://icml.cc/virtual/2026/poster/65791

arXiv：https://arxiv.org/abs/2603.02630

作者：Zhi Hong ⋅ Qian Zhang ⋅ Jiahang Sun ⋅ Zhiwei Shang ⋅ Mingze Kong ⋅ Xiangyi Wang ⋅ Yao Shu ⋅ Zhongxiang Dai

关键词：多智能体，GNN

12 Navigating the Energy Landscape of Collaboration: Multi-Agent Communication Graph Generation via Score-Based Diffusion

链接：https://icml.cc/virtual/2026/poster/61913

作者：GuanHao Zhao ⋅ Wenbo Lu ⋅ Cheng Cheng ⋅ Zhenya Huang ⋅ Wei Song ⋅ Zhiding Liu ⋅ Runze Wu ⋅ Enhong Chen

关键词：多智能体通信，通信拓扑图

13 GraphFlow: A Graph-Based Workflow Management for Efficient LLM-Agent Serving

链接：https://icml.cc/virtual/2026/poster/66432

作者：Ao Li ⋅ Shangpeng Yang ⋅ Fahao Chen ⋅ Tianheng Xu ⋅ Peng Li ⋅ su zhou

关键词：基于图的Agent服务

14 Factored Value Functions for Graph-Based Multi-Agent Reinforcement Learning

链接：https://icml.cc/virtual/2026/poster/62462

arXiv：http://arxiv.org/abs/2601.11401v1

作者：Ahmed Rashwan ⋅ Keith Briggs ⋅ Chris Budd ⋅ Lisa Kreusser

关键词：基于图多智能体强化学习

15 Embodied Task Planning via Graph-Informed Action Generation with Large Lanaguage Model

链接：https://icml.cc/virtual/2026/poster/61431

arXiv：https://arxiv.org/abs/2601.21841

作者：Xiang Li ⋅ Ning Yan ⋅ Masood Mortazavi

关键词：具身智能体，图结构，LLM

16 SAGE-NAS: Synergizing LLM-Based Semantic Agent with Graph-Based Evaluator for Neural Architecture Search

链接：https://icml.cc/virtual/2026/poster/62021

作者：Kaiqi Lin ⋅ Jianping Luo

关键词：神经网络架构搜索，Agent

17 Weaving Graph over Tokens: Contextualizing Structured Sequences for LLMs

链接：https://icml.cc/virtual/2026/poster/62437

作者：Jiaxuan Chen ⋅ Zixing Zhang ⋅ Ruijun Mao ⋅ Wei Sun ⋅ Zhicheng Liang ⋅ Yuhang Zhang ⋅ Yaxi Liu ⋅ Fangxin Wang

关键词：生成式图语言模型

18 D${}^3$: Dynamic Directional Graph-Constrained Data Scheduling for LLM Training

链接：https://icml.cc/virtual/2026/poster/62470

作者：Xu Yuanjian ⋅ Jianing Hao ⋅ Guang Zhang ⋅ Zhong Li

关键词：数据调度，图结构建模

19 GAUSS: Graph-Assisted Uncertainty Quantification using Structure and Semantics for Long-Form Generation in LLMs

链接：https://icml.cc/virtual/2026/poster/65373

作者：Karthik Somayaji NS ⋅ Yuxuan Yin ⋅ Peng Li

关键词：长文本生成，图辅助的不确定性量化

20 Memory is Reconstructed, Not Retrieved: Graph Memory for LLM Agents

链接：https://icml.cc/virtual/2026/poster/60697

作者：Shuo Ji ⋅ yibo li ⋅ Bryan Hooi

关键词：智能体记忆，Agent