Lightweight reranking for language model generations

TL;DR

This paper introduces a low-overhead reranking method for LLM outputs that improves generation quality across multiple tasks by leveraging pairwise statistics, with theoretical analysis and empirical validation.

Contribution

It presents a novel reranking approach based on pairwise statistics that requires minimal compute and no additional training, extending self-consistency for better selection of top generations.

Findings

Significant improvements in code generation quality.

Robust enhancements in autoformalization, summarization, and translation.

Additional token probability access further boosts performance.

Abstract

Large Language Models (LLMs) can exhibit considerable variation in the quality of their sampled outputs. Reranking and selecting the best generation from the sampled set is a popular way of obtaining strong gains in generation quality. In this paper, we present a novel approach for reranking LLM generations. Unlike other techniques that might involve additional inferences or training a specialized reranker, our approach relies on easy to compute pairwise statistics between the generations that have minimal compute overhead. We show that our approach can be formalized as an extension of self-consistency and analyze its performance in that framework, theoretically as well as via simulations. We show strong improvements for selecting the best k generations for code generation tasks as well as robust improvements for the best generation for the tasks of autoformalization, summarization, and translation. While our approach only assumes black-box access to LLMs, we show that additional access to token probabilities can improve performance even further.