Activity (Process)

Best-of-N Sampling (BoN Sampling)

Best-of-N (BoN) sampling is a technique where a model generates multiple, or 'N', alternative outputs, and a reward model scores them to select the best one. While commonly used for inference-time alignment through reranking, the core mechanism of BoN sampling can also be adapted for training purposes, such as in rejection sampling.

0

1

Updated 2026-05-03

Contributors are:

Who are from:

Tags

Ch.5 Inference - Foundations of Large Language Models

Foundations of Large Language Models

Foundations of Large Language Models Course

Computing Sciences

Ch.4 Alignment - Foundations of Large Language Models