Graph-R1

Graph-R1 copied to clipboard

作者您好！这篇文章将GRPO算法应用到了知识图谱+RAG领域非常的令人瞩目，但是如果我们复现的时候手边没有大现存的A100（80G）显卡，比较难以复现代码段中的强化学习部分，同时又很想直接推理结果。请问您是否方便直接提供强化学习完成之后的模型权重？

shiqiuchangatp

Corpus构建

8

非常棒的工作！ 关于Graph-R1， 我还有如下问题想请教：请问每个数据集的corpus是如何构建的？是基于sample出来的 train&test set的content的构建的还是基于原始数据集的content？另外如果我想使用论文中未用到的数据集来训练，该如何构建corpus呢？

FanDongzhe123

如何对构建的超图进行可视化？有没有推荐的工具？

4

非常棒的工作！！！ 我还有几个问题想要请教您： 1.GraphR1论文中的超图构建是否完全基于您上一篇论文的HyperGraphRAG？我计划只对HyperGraphRAG进行一些测试，我看您上一篇论文HyperGraphRAG有更多的细节。我是否可以参考HyperGraphRAG论文，使用本文的超图构建方式进行构建？ 2.还有如何对expr下面构建出来的超图文件进行可视化，有无推荐的可视化工具？ 期待您的回复！ 万分感谢！

Ceaserk01

有关vllm和flash-attn安装问题

1

作者您好！ 这是一份很赞的工作，但是我在部署时遇到了一些问题，具体如下两个bug： 希望能得到您的帮助！

lazyloafer

知识检索

1

作者您好，我注意到当前的代码实现主要采用了向量检索方式，而未引入图检索机制。考虑到图结构在表达实体关系和上下文方面具有天然优势，图检索在语义推理和结构化信息扩展方面可能更具潜力。请问在设计 Graph-R1 时，出于哪些考虑未采用图检索？是否有相关的性能或复杂度权衡？另外，若希望实现图检索，在这个项目上实现可行吗?

ninjaX2o

Question About Dataset Construction Cost

2

Hi, thank you for sharing your impressive work on Graph-R1! I was impressed by the reported efficiency for constructing the graph. Could you share an estimate of the total cost...

ninjaX2o

wall-clock time / GPU-hours to reproduce (GRPO on Qwen2.5-3B/7B)

1

Hi, thanks for releasing the code and paper! Could you share a ballpark training duration so we can better budget the costs that would be necessary to reproduce those results?...

mrT333

So does this need retraining on each new dataset?

3

This looks interesting, I have been working with graphRAG for some time now and feel I am getting pretty great results but always looking for more! It looks like this...

voycey

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

1

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...

deeplearning-LI

How are the hypergraph-related JSON files generated?

3

Hi, thank you for sharing this great project! I’m currently studying the codebase and noticed that the script relies on the following JSON files: kv_store_text_chunks.json kv_store_entities.json kv_store_hyperedges.json Could you kindly...

luckyyangrun