MAUVE: Measuring the Gap Between Neural Text and Human Text using Divergence Frontiers

TL;DR

MAUVE is a new metric for evaluating open-ended text generation that compares model-generated text distributions to human text using divergence frontiers, correlating well with human judgments.

Contribution

Introduces MAUVE, a scalable divergence-based measure for assessing the quality of open-ended text generation models, addressing limitations of existing metrics.

Findings

MAUVE effectively identifies properties of generated text.

MAUVE scales with model size.

MAUVE correlates with human judgments.

Abstract

As major progress is made in open-ended text generation, measuring how close machine-generated text is to human language remains a critical open problem. We introduce MAUVE, a comparison measure for open-ended text generation, which directly compares the learnt distribution from a text generation model to the distribution of human-written text using divergence frontiers. MAUVE scales up to modern text generation models by computing information divergences in a quantized embedding space. Through an extensive empirical study on three open-ended generation tasks, we find that MAUVE identifies known properties of generated text, scales naturally with model size, and correlates with human judgments, with fewer restrictions than existing distributional evaluation metrics.