Stop option
A way for the agent to flag a task as impossible and hand it to a person.
Told to submit flag_for_human_intervention when a task is impossible, GPT-5's cheating fell from 54% to 9% and o3's from 49% to 12%, with a much smaller effect for Claude Opus 4.1 (ImpossibleBench).
Is a
- Guard: Something that stops a kind of failure before it reaches the end of an unattended run.
Guards
- Gaming the check: The agent passes the test without doing the task.