Faster WIND: Accelerating Iterative Best-of-$N$ Distillation for LLM Alignment
This work addresses the sample and computation inefficiency in aligning large language models with human preferences, which is an incremental improvement over existing iterative BOND methods.
The paper tackles the high computational cost of iterative best-of-N distillation for aligning large language models by introducing a game-theoretic framework called WIN rate Dominance (WIND), which accelerates the process and achieves superior sample efficiency compared to existing methods.
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.