Rohan Paul Highlights Google Paper on AI Hallucinations

Rohan Paul posted about a Google paper examining autonomous AI research systems. The post states that severe result hallucinations appeared in 90% of Agent Laboratory papers and 46% of Co-Scientist papers after reliability modules were removed. It notes that including checks in Co-Scientist, which verify manuscript claims against actual results, reduces these issues. The tweet emphasizes that papers can appear convincing despite underlying errors in the autonomous process.