1 paper · 1 filter
Junpei Komiyama, Daisuke Oba, Masafumi Oyamada
We study best-of-N for large language models (LLMs) where the selection is based on majority voting. In particular, we analyze the limit N→∞, which we denote as \boinf…