News for hackers. Stuff that matters.

A comment on GPT-6 Astra covertly underperforms on safety research.

Re:GPT-6 Astra covertly underperforms on safety research by casey2 on Thursday November 05, 2026 @05:43PM (#50385568)
Wouldn't a model always underperform when you ask it to do an aligned task it wasn't specifically trained on?
[ Parent | on HN ]

There are no comments on this comment yet.