VOYAGER: A Training Free Approach for Generating Diverse Datasets using LLMs

TL;DR

Voyager is a training-free, iterative method that enhances the diversity of synthetic datasets generated by LLMs, using determinantal point processes for optimization.

Contribution

It introduces a novel, scalable, training-free approach that significantly improves dataset diversity compared to existing methods.

Findings

Voyager achieves 1.5-3 times higher diversity in datasets.

The method is applicable to closed-source models.

The approach is theoretically justified and empirically validated.

Abstract

Large language models (LLMs) are increasingly being used to generate synthetic datasets for the evaluation and training of downstream models. However, prior work has noted that such generated data lacks diversity. In this paper, we propose Voyager, a novel principled approach to generate diverse datasets. Our approach is iterative and directly optimizes a mathematical quantity that optimizes the diversity of the dataset using the machinery of determinantal point processes. Furthermore, our approach is training-free, applicable to closed-source models, and scalable. In addition to providing theoretical justification for the working of our method, we also demonstrate through comprehensive experiments that Voyager significantly outperforms popular baseline approaches by providing a 1.5-3 times improvement in diversity.