If GPT-5 is considered best aligned by this metric, I am highly skeptical that the metric is measuring any general sense of alignment. I get why GPT-5 barely does anything except follow instructions…

Posters: repligate. repligate

2 likes recorded by the source corpus.

Topics: AI & LLMs, AI safety

View Post: https://x.com/repligate/status/1972734481371971838