TextGAIL: Generative Adversarial Imitation Learning for Text Generation

TL;DR

TextGAIL introduces a novel adversarial imitation learning framework utilizing pre-trained language models and contrastive discriminators to enhance the quality and diversity of generated text, outperforming traditional GAN approaches.

Contribution

The paper proposes a new text generation method combining imitation learning, contrastive discriminators, and PPO, addressing GAN limitations in text generation.

Findings

TextGAIL outperforms MLE baselines in quality and diversity.

Discriminator effectively provides reliable reward signals.

Method is validated on various unconditional and conditional tasks.

Abstract

Generative Adversarial Networks (GANs) for text generation have recently received many criticisms, as they perform worse than their MLE counterparts. We suspect previous text GANs' inferior performance is due to the lack of a reliable guiding signal in their discriminators. To address this problem, we propose a generative adversarial imitation learning framework for text generation that uses large pre-trained language models to provide more reliable reward guidance. Our approach uses contrastive discriminator, and proximal policy optimization (PPO) to stabilize and improve text generation performance. For evaluation, we conduct experiments on a diverse set of unconditional and conditional text generation tasks. Experimental results show that TextGAIL achieves better performance in terms of both quality and diversity than the MLE baseline. We also validate our intuition that TextGAIL's discriminator demonstrates the capability of providing reasonable rewards with an additional task.