alt Hacker NewsThe LLM needs to read the CoT as part of the conversation. You can ask the models to share them with you. Stronger models will refuse, while weaker ones can be “jailbroken”.
The LLM needs to read the CoT as part of the conversation. You can ask the models to share them with you. Stronger models will refuse, while weaker ones can be “jailbroken”.
I don’t think that answers what I am wondering. Asked differently why is encoding/decoding the cot the concern of the LLM?