RSA-Control: A Pragmatics-Grounded Lightweight Controllable Text Generation Framework

TL;DR

RSA-Control is a training-free framework for controllable text generation that uses pragmatic reasoning between speakers and listeners, with an adjustable control strength, to produce attribute-controlled, fluent, and consistent texts.

Contribution

It introduces a novel pragmatics-grounded, training-free approach with an adjustable parameter for improved controllable text generation.

Findings

Achieves strong attribute control in generated texts.

Maintains language fluency and content consistency.

Effective across different task types and language models.

Abstract

Despite significant advancements in natural language generation, controlling language models to produce texts with desired attributes remains a formidable challenge. In this work, we introduce RSA-Control, a training-free controllable text generation framework grounded in pragmatics. RSA-Control directs the generation process by recursively reasoning between imaginary speakers and listeners, enhancing the likelihood that target attributes are correctly interpreted by listeners amidst distractors. Additionally, we introduce a self-adjustable rationality parameter, which allows for automatic adjustment of control strength based on context. Our experiments, conducted with two task types and two types of language models, demonstrate that RSA-Control achieves strong attribute control while maintaining language fluency and content consistency. Our code is available at https://github.com/Ewanwong/RSA-Control.