Controlled Text Generation for Large Language Model with Dynamic Attribute Graphs
This work addresses the problem of generating text with specific attributes for users of large language models, representing an incremental advance in controlled text generation methods.
The paper tackles controlled text generation for large language models by introducing a pluggable framework called DATG that uses dynamic attribute graphs to modulate key attribute words, achieving up to 19.29% improvement in control accuracy and reduced perplexity across toxicity mitigation and sentiment transformation tasks.
Controlled Text Generation (CTG) aims to produce texts that exhibit specific desired attributes. In this study, we introduce a pluggable CTG framework for Large Language Models (LLMs) named Dynamic Attribute Graphs-based controlled text generation (DATG). This framework utilizes an attribute scorer to evaluate the attributes of sentences generated by LLMs and constructs dynamic attribute graphs. DATG modulates the occurrence of key attribute words and key anti-attribute words, achieving effective attribute control without compromising the original capabilities of the model. We conduct experiments across four datasets in two tasks: toxicity mitigation and sentiment transformation, employing five LLMs as foundational models. Our findings highlight a remarkable enhancement in control accuracy, achieving a peak improvement of 19.29% over baseline methods in the most favorable task across four datasets. Additionally, we observe a significant decrease in perplexity, markedly improving text fluency.