Concept · 5 episode(s)

Parallel Sampling

← all concepts

Definition

Parallel sampling generates many candidate responses from a model at once and then picks among them — by vote, by verifier, by reward model. It’s a simple way to trade inference compute for quality and the underlying mechanism of pass@k and self-consistency.

Episodes covering this

Worth reading next

Papers we haven't done a deep dive on yet, but would recommend on this topic.