Generating Symbolic Reasoning Problems with Transformer GANs

TL;DR

This paper explores the use of Transformer-based GANs to generate high-quality, challenging symbolic reasoning problems, demonstrating their utility in training classifiers and modifying data difficulty.

Contribution

It introduces Transformer GANs for generating symbolic reasoning data, capable of producing syntactically correct instances without autoregression, and shows how to manipulate data difficulty via generator objectives.

Findings

Generated data can replace real training data effectively.

GANs produce syntactically correct symbolic instances without autoregression.

Adding classifier uncertainty makes generated data more challenging for classifiers.

Abstract

We study the capabilities of GANs and Wasserstein GANs equipped with Transformer encoders to generate sensible and challenging training data for symbolic reasoning domains. We conduct experiments on two problem domains where Transformers have been successfully applied recently: symbolic mathematics and temporal specifications in verification. Even without autoregression, our GAN models produce syntactically correct instances. We show that the generated data can be used as a substitute for real training data when training a classifier, and, especially, that training data can be generated from a dataset that is too small to be trained on directly. Using a GAN setting also allows us to alter the target distribution: We show that by adding a classifier uncertainty part to the generator objective, we obtain a dataset that is even harder to solve for a temporal logic classifier than our original dataset.