natufunu·4 か月前·議論Great post! I wonder why MCTS is not more popular as a test time compute harness. Did you compare performance of MCTS (without distillation) against other methods (eg best of N) with the same compute budget?
natufunu·6 か月前·議論One thing I learned from this is that my prompts are much less detailed than what author has been using.Very cool work and great write up.