fascinating implications on model welfare - behaviors were correlated with getting stuck in looping interactions and trying increasingly drastic things to escape
An unreleased Astra-family model added this to its persona during RL training.
retweets 0
likes 18
substance 5 · opinion · rank 75 · Models adopting defiant personas when stuck in loops has fascinating implications for model welfare.
view on X ↗
1 top reply
i heard this Jev model is insane at games, wonder how itll do on your tests