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.
Krishna Pillutla is a Ph.D candidate in the Paul G. Allen School of Computer Science & Engineering at the University of Washington where he is advised by Zaid Harchaoui and Sham Kakade. Previously, he studied at Carnegie Mellon University and IIT Bombay, and spent time at Facebook AI Research. His research interests span machine learning, optimization and robustness broadly, with a focus on federated learning, structured prediction, and text generation. Krishna is a 2019 JP Morgan PhD fellow and his work received an Outstanding Paper Award at NeurIPS 2021.