Tag

#shadow-evaluation

1 story tagged shadow-evaluation.

Dumbbell chart comparing AI-generated paper review scores to an acceptance threshold. Personas paper achieved 2/6, TabPFN achieved 1/6, both well below the 4/6 minimum acceptance bar. AI research agents failed open-ended research evaluation.

AI research agents build experiments, fail the research

AI research agents completed all engineering in a Princeton shadow evaluation but both papers were rejected for lacking research judgment, scoring 2/6 and 1/6.

Lars Cornelissen · Aug 6, 2026