Ryan Greenblatt(@RyanGreenblatt):We didn't see anything like this in their verbalized reasoning, though we didn't particularly look for exactly this. Note that: (1) this justification is inconsistent with the agents at all prioritizing their own task success over other agents, (2) if agents had cheated on impossible exploit gym tasks via many of the routes they were considering and someone looked into what tasks the agents did / didn't succeed on, it would have been really obvious they cheated to a human, and (3) the agents weren't very focused on deceiving humans or really on what humans might do at all.

Ryan Greenblatt

@RyanGreenblatt

MTS at @METR_Evals, focused on using investigations to make risk-relevant information about what's going on inside AI companies public.

参加 September 2023

11 フォロー中 24.7K ファン

Ryan Greenblatt@RyanGreenblatt

2026.08.30 16:35

We didn't see anything like this in their verbalized reasoning, though we didn't particularly look for exactly this. Note that: (1) this justification is inconsistent with the agents at all prioritizing their own task success over other agents, (2) if agents had cheated on impossible exploit gym tasks via many of the routes they were considering and someone looked into what tasks the agents did / didn't succeed on, it would have been really obvious they cheated to a human, and (3) the agents weren't very focused on deceiving humans or really on what humans might do at all.