FedPref enables collaborative AI for radiology report extraction without data sharing

Researchers have developed FedPref, a novel federated learning approach designed to improve structured radiology report extraction. This method allows institutions with limited or unevenly distributed data to collaborate by training compact adapters for the Qwen3_8B model. FedPref enhances extraction accuracy, particularly for smaller sites, by using frozen public language models to propose extractions and local annotations to rank them, sharing only model updates rather than raw data. AI

IMPACT This method could enable more effective AI-driven analysis of medical reports across institutions with varying data availability.

RANK_REASON The item describes a new research paper detailing a novel method for AI-based data extraction. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• alphaXiv
• arXiv
• CatalyzeX
• DagsHub
• FedPref
• Gotit.pub
• Hugging Face
• Qwen3_8B
• ScienceCast

• paper
• model release
• infra

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