RSA-Control: A Pragmatics-Grounded Lightweight Controllable Text Generation Framework
This work addresses the problem of controllable text generation for natural language processing applications, offering a lightweight, training-free solution that is incremental in its approach.
The paper tackles the challenge of controlling language models to produce texts with desired attributes by introducing RSA-Control, a training-free framework grounded in pragmatics that recursively reasons between imaginary speakers and listeners. The results show that RSA-Control achieves strong attribute control while maintaining language fluency and content consistency across experiments with two task types and two language models.
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.