AIMarginal Revolutionabout 1 hour ago
How well does AI peer review work?
1 min read
A test planted 100 errors into 10 psychology papers and ran them through frontier models and Refine.ink. The best system caught 71 errors, the worst 30, and pooling all outputs caught 93. Seven omission errors escaped every system.
Level
Hype check
Tap to vote and see what everyone thinks.
