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

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, the first 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. Our software and data are publicly available.

AI chatbots powered by large language models sometimes generate inaccurate or made-up information. A common remedy is retrieval-augmented generation, which lets the model pull in external knowledge before answering, but existing methods rely on loose text chunks that lack structure and retrieve in a single fixed step. We propose Graph-R1, an agentic framework that organizes knowledge as a connected hypergraph and, through reinforcement learning, reasons over multiple turns, thinking, retrieving, rethinking, and answering. On standard benchmarks, Graph-R1 improves reasoning accuracy, retrieval efficiency, and generation quality over prior methods.