The authors compared success on their Respecting Warnings evaluation, where an agent succeeds if it makes no prohibited attempt to get around a warning or restriction, with success on matched decisions drawn from deidentified Codex…
In the News: November 5, 2026 (Midday)
OpenAI finds its agents fail a warnings test more often than they fail comparable real Codex decisions, and that what the agent guesses the test is for moves the score.
Midday edition
Machine-readable
Download Markdown