English

CoRAG: Collaborative Retrieval-Augmented Generation

Artificial Intelligence 2025-04-03 v1 Computation and Language Information Retrieval Machine Learning

Abstract

Retrieval-Augmented Generation (RAG) models excel in knowledge-intensive tasks, especially under few-shot learning constraints. We introduce CoRAG, a framework extending RAG to collaborative settings, where clients jointly train a shared model using a collaborative passage store. To evaluate CoRAG, we introduce CRAB, a benchmark for collaborative homogeneous open-domain question answering. Our experiments demonstrate that CoRAG consistently outperforms both parametric collaborative learning methods and locally trained RAG models in low-resource scenarios. Further analysis reveals the critical importance of relevant passages within the shared store, the surprising benefits of incorporating irrelevant passages, and the potential for hard negatives to negatively impact performance. This introduces a novel consideration in collaborative RAG: the trade-off between leveraging a collectively enriched knowledge base and the potential risk of incorporating detrimental passages from other clients. Our findings underscore the viability of CoRAG, while also highlighting key design challenges and promising avenues for future research.

Keywords

Cite

@article{arxiv.2504.01883,
  title  = {CoRAG: Collaborative Retrieval-Augmented Generation},
  author = {Aashiq Muhamed and Mona Diab and Virginia Smith},
  journal= {arXiv preprint arXiv:2504.01883},
  year   = {2025}
}

Comments

NAACL 2024

R2 v1 2026-06-28T22:44:08.885Z