Reasoning with Sampling: Your Base Model is Smarter Than You Think

TL;DR

We find a training-free sampling algorithm that achieves reasoning boosts on base models comparable to those obtained by RL techniques.

Opening excerpt from the authors’ abstract. source