An OpenAI-led study finds current AI models struggle to manipulate their reasoning, but warns that more advanced systems could eventually obscure decision processes during monitoring ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results