Why not compare with iterative or multi-round retrieval works, since one-time interaction is discussed in the paper?

[deeplearning-LI]

From the perspective of the overall workflow, Graph-R1 seems to perform step-by-step decomposition of queries and graph retrieval. What advantages does this offer compared to iterative RAG? (Does it have advantages in terms of performance?)

Thank you for your answers to the questions.

[LHRLAB]

Thank you for your question.
While Search-R1 and R1-Searcher are multi-turn RAG methods, Graph-R1 uses step-by-step reasoning with fine-grained hypergraph representation, which help improve retrieval efficiency, reasoning accuracy, and generation quality.