FedMosaic framework enhances federated RAG with parametric adapters

Researchers have introduced FedMosaic, a novel framework for federated retrieval-augmented generation (FedRAG) that addresses the challenges of privacy-aware domains. Unlike traditional RAG, FedMosaic uses parametric adapters to encode documents, preventing the exchange of raw text. The system clusters documents into multi-document adapters with specific masks to reduce storage and communication overhead, while also employing selective adapter aggregation to combine only relevant and non-conflicting adapters. Experiments demonstrate FedMosaic's superior accuracy and significant cost reductions compared to existing methods. AI

IMPACT This research could enable more private and efficient deployment of RAG systems in sensitive domains.

RANK_REASON The cluster contains a research paper detailing a new framework for federated retrieval-augmented generation. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CL →

• alphaXiv
• arXiv
• Boyi Liu
• CatalyzeX
• DagsHub
• FedMosaic
• FedRAG
• Gotit.pub
• Hugging Face
• large-language models
• retrieval-augmented generation
• ScienceCast

• paper
• infra

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →