EIN-RECAP: Evidence-Grounded Multi-Agent Summarization and Retrieval of Infectious-Disease Expert Discussions

EIN-RECAP, an AI platform for EIN infectious-disease expert discussions, improved coverage on nine high-disagreement threads, but experts preferred a simpler pipeline over a multi-agent workflow.

• Why it matters: Clinicians need to find relevant, evolving observations across years of discussions that may not yet appear in published evidence. Automated summarization and retrieval can miss these nuances, so expert feedback is needed to validate outputs.
• What they did: The team built EIN-RECAP with retrieval-grounded summarization, citation validation, and hybrid search, then evaluated methods on 3,185 threads and ran blinded expert review on a recent subset. A multi-agent workflow scored higher automatically, and expert reviewers assessed quality and issues.
• The result: An ensemble approach improved discussion coverage and key clinical points on nine high-disagreement threads, but retrieval relevance remained a limitation and experts flagged repetition, inaccuracies, and overconfident recommendations in the more complex system.

AI summary of the abstract · may contain errors · not medical advice · read the paper.