Which Discriminator for Cooperative Text Generation?

TL;DR

This paper compares three types of transformer-based discriminators—bidirectional, left-to-right, and generative—for guiding cooperative text generation, analyzing their accuracy, sample quality, and computational efficiency.

Contribution

It provides a comprehensive evaluation of different discriminator architectures for cooperative decoding in language models, including implementation details of Monte Carlo Tree Search.

Findings

Bidirectional discriminators offer high accuracy but are computationally intensive.

Left-to-right discriminators balance quality and efficiency effectively.

Generative discriminators facilitate seamless integration with language models.

Abstract

Language models generate texts by successively predicting probability distributions for next tokens given past ones. A growing field of interest tries to leverage external information in the decoding process so that the generated texts have desired properties, such as being more natural, non toxic, faithful, or having a specific writing style. A solution is to use a classifier at each generation step, resulting in a cooperative environment where the classifier guides the decoding of the language model distribution towards relevant texts for the task at hand. In this paper, we examine three families of (transformer-based) discriminators for this specific task of cooperative decoding: bidirectional, left-to-right and generative ones. We evaluate the pros and cons of these different types of discriminators for cooperative generation, exploring respective accuracy on classification tasks along with their impact on the resulting sample quality and computational performances. We also provide the code of a batched implementation of the powerful cooperative decoding strategy used for our experiments, the Monte Carlo Tree Search, working with each discriminator for Natural Language Generation.