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

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Luo, Haoran, E, Haihong, Chen, Guanting, Lin, Qika, Guo, Yikai, Xu, Fangzhi, Kuang, Zemin, Song, Meina, Wu, Xiaobao, Zhu, Yifan, Tuan, Luu Anh
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912508940386304
author Luo, Haoran
E, Haihong
Chen, Guanting
Lin, Qika
Guo, Yikai
Xu, Fangzhi
Kuang, Zemin
Song, Meina
Wu, Xiaobao
Zhu, Yifan
Tuan, Luu Anh
author_facet Luo, Haoran
E, Haihong
Chen, Guanting
Lin, Qika
Guo, Yikai
Xu, Fangzhi
Kuang, Zemin
Song, Meina
Wu, Xiaobao
Zhu, Yifan
Tuan, Luu Anh
contents Retrieval-Augmented Generation (RAG) mitigates hallucination in LLMs by incorporating external knowledge, but relies on chunk-based retrieval that lacks structural semantics. GraphRAG methods improve RAG by modeling knowledge as entity-relation graphs, but still face challenges in high construction cost, fixed one-time retrieval, and reliance on long-context reasoning and prompt design. To address these challenges, we propose Graph-R1, an agentic GraphRAG framework via end-to-end reinforcement learning (RL). It introduces lightweight knowledge hypergraph construction, models retrieval as a multi-turn agent-environment interaction, and optimizes the agent process via an end-to-end reward mechanism. Experiments on standard RAG datasets show that Graph-R1 outperforms traditional GraphRAG and RL-enhanced RAG methods in reasoning accuracy, retrieval efficiency, and generation quality.
format Preprint
id arxiv_https___arxiv_org_abs_2507_21892
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Graph-R1: Towards Agentic GraphRAG Framework via End-to-end Reinforcement Learning
Luo, Haoran
E, Haihong
Chen, Guanting
Lin, Qika
Guo, Yikai
Xu, Fangzhi
Kuang, Zemin
Song, Meina
Wu, Xiaobao
Zhu, Yifan
Tuan, Luu Anh
Computation and Language
Retrieval-Augmented Generation (RAG) mitigates hallucination in LLMs by incorporating external knowledge, but relies on chunk-based retrieval that lacks structural semantics. GraphRAG methods improve RAG by modeling knowledge as entity-relation graphs, but still face challenges in high construction cost, fixed one-time retrieval, and reliance on long-context reasoning and prompt design. To address these challenges, we propose Graph-R1, an agentic GraphRAG framework via end-to-end reinforcement learning (RL). It introduces lightweight knowledge hypergraph construction, models retrieval as a multi-turn agent-environment interaction, and optimizes the agent process via an end-to-end reward mechanism. Experiments on standard RAG datasets show that Graph-R1 outperforms traditional GraphRAG and RL-enhanced RAG methods in reasoning accuracy, retrieval efficiency, and generation quality.
title Graph-R1: Towards Agentic GraphRAG Framework via End-to-end Reinforcement Learning
topic Computation and Language
url https://arxiv.org/abs/2507.21892