Right, and this is what "reasoning LLMs" work around by having explicitly labelly "reasoning tokens". This lets them "save" the plan between tokens, so when re

Right, and this is what "reasoning LLMs" work around by having explicitly labelly "reasoning tokens". This lets them "save" the plan between tokens, so when regenerating the new token it is following the plan.