Plug and Play with Prompts: A Prompt Tuning Approach for Controlling Text Generation

TL;DR

This paper introduces a prompt tuning method that enables controlled text generation in language models using minimal data, effectively steering outputs and reducing harmful content.

Contribution

It presents a novel prompt tuning approach trained with small datasets to control language generation and mitigate toxicity in outputs.

Findings

Effective control over generated text using prompt embeddings.

Works with very limited training data (a few hundred examples).

Reduces toxic and biased language in generated outputs.

Abstract

Transformer-based Large Language Models (LLMs) have shown exceptional language generation capabilities in response to text-based prompts. However, controlling the direction of generation via textual prompts has been challenging, especially with smaller models. In this work, we explore the use of Prompt Tuning to achieve controlled language generation. Generated text is steered using prompt embeddings, which are trained using a small language model, used as a discriminator. Moreover, we demonstrate that these prompt embeddings can be trained with a very small dataset, with as low as a few hundred training examples. Our method thus offers a data and parameter efficient solution towards controlling language model outputs. We carry out extensive evaluation on four datasets: SST-5 and Yelp (sentiment analysis), GYAFC (formality) and JIGSAW (toxic language). Finally, we demonstrate the efficacy of our method towards mitigating harmful, toxic, and biased text generated by language models.