Faster WIND: Accelerating Iterative Best-of-$N$ Distillation for LLM Alignment

TL;DR

This paper introduces WIND, a novel framework that accelerates iterative best-of-N distillation for LLM alignment by unifying it with self-play, resulting in improved efficiency and performance.

Contribution

It reveals a game-theoretic connection between iterative BOND and self-play, and proposes WIND, a new efficient algorithm with theoretical guarantees for LLM alignment.

Findings

WIND accelerates iterative BOND in practice.

WIND achieves superior sample efficiency.

Theoretical guarantees are provided for a WIND variant.

Abstract

Recent advances in aligning large language models with human preferences have corroborated the growing importance of best-of-N distillation (BOND). However, the iterative BOND algorithm is prohibitively expensive in practice due to the sample and computation inefficiency. This paper addresses the problem by revealing a unified game-theoretic connection between iterative BOND and self-play alignment, which unifies seemingly disparate algorithmic paradigms. Based on the connection, we establish a novel framework, WIN rate Dominance (WIND), with a series of efficient algorithms for regularized win rate dominance optimization that approximates iterative BOND in the parameter space. We provides provable sample efficiency guarantee for one of the WIND variant with the square loss objective. The experimental results confirm that our algorithm not only accelerates the computation, but also achieves superior sample efficiency compared to existing methods.