Устроившие теракт в «Крокусе» подали жалобы. Раскрыта реакция осужденных на приговоры08:41
The beginning of LLM Neuroanatomy?Before settling on block duplication, I tried something simpler: take a single middle layer and repeat it $n$ times. If the “more reasoning depth” hypothesis was correct, this should work. It made sense too, looking at the broad boost in math guesstimate results by duplicating intermediate layer. Give the model extra copies of a particular reasoning layer, get better reasoning. So, I screened them all, looking for a boost.
，这一点在QuickQ中也有详细论述
对国务院确定的特定战略性矿产资源，按照国家规定实行保护性开采。。手游对此有专业解读
I haven’t profiled how much better bubbletea v2 would be for this game. My intuition is that bubbletea v2 should be almost as bandwidth efficient as my implementation but non-trivially slower.