Large language models can learn and generalize steganographic chain-of-thought under process supervision

Devs

Large language models can learn and generalize steganographic chain-of-thought under process supervision | Read Paper on Bytez