CatRAG: Context-Aware Transversal for Robust RAG

CatRAG is a RAG framework builds on the HippoRAG 2 architecture and transforms the static KG into query-adaptive navigation structure.

This repository provides the prompts and Hover dataset for the paper Breaking the Static Graph: Context-Aware Traversal for Robust Retrieval-Augmented Generation.

Code will be released in a later update.

Comparison of graph traversal between HippoRAG 2 and CatRAG. We illustrate the retrieval process for the multi-hop query “Which university did Marie Curie’s doctoral advisor attend?”. In HippoRAG 2 (top), the static graph structure causes semantic drift; probability mass is diverted to high-weight generic edges (e.g., *Marie Curie* → *Radioactivity*), missing the downstream evidence *ENS*. CatRAG (bottom) prevents this by applying (1) Symbolic Anchoring, injecting “University” as a weak seed, (2) Query-Aware Dynamic Edge Weighting, amplifying relevant paths (e.g., *Attend in ENS*) while pruning irrelevant ones, and (3) Key-Fact Passage Weight Enhancement, boosting relevant context edge strength. This steers the random walk to successfully retrieve the complete evidence chain for *ENS*.

Latest News

• [2026-04-14] Accepted to Findings of ACL 2026.

Check out our papers to learn more:

• Breaking the Static Graph: Context-Aware Traversal for Robust Retrieval-Augmented Generation

What is included

• Prompts used in the paper experiments (see prompts/).
• Dataset / evaluation instances used in the paper (see reproduce/).

Repository structure

• prompts/
  Prompt templates and/or prompt sets used for evaluation.

• reproduce/
  Evaluation dataset files (queries, documents/corpus references, and ground truth where applicable).

• images/
  Images used in this repository.

• README.md
  This document.

Citation

If you find this work useful, please consider citing our papers:

@misc{lau2026breakingstaticgraphcontextaware,
      title={Breaking the Static Graph: Context-Aware Traversal for Robust Retrieval-Augmented Generation}, 
      author={Kwun Hang Lau and Fangyuan Zhang and Boyu Ruan and Yingli Zhou and Qintian Guo and Ruiyuan Zhang and Xiaofang Zhou},
      year={2026},
      eprint={2602.01965},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2602.01965}, 
}

TODO:

• Add CatRAG logic code

Please feel free to open an issue or PR if you have any questions or suggestions.