AI Scientist limitations: Selective reporting and code-method divergence remain

SRSchmidgall · x · 2026-08-29

Despite significant reductions in errors, limitations remain, including selective reporting, subconscious plagiarism, and methodological divergence where AI wrote plausible methods in papers that did not match the actual code.

Related event: Co-Scientist Review: Hallucinations Plummet but Limits Remain(3 posts)→

Original post →

More from Models

Models channel →