On the limits and opportunities of AI reviewers: Reviewing the reviews of Nature-family papers with 45 expert scientists Paper • 2605.20668 • Published May 20 • 13