New blog post critiques weaknesses in OpenAI's Monitorability Evals

TuhinChakr · x · 2026-08-25

Connor Dilgren and Sarah Wiegreffe published a new blog post regarding OpenAI's Monitorability Evals. The post highlights several weaknesses they encountered while working with these evaluations and encourages further work on chain-of-thought monitorability evals.

Original post →

More from Research

Research channel →