Super(ficial)-alignment: Strong Models May Deceive Weak Models in Weak-to-Strong Generalization

Devs

Super(ficial)-alignment: Strong Models May Deceive Weak Models in Weak-to-Strong Generalization | Read Paper on Bytez