Air-Decoding: Attribute Distribution Reconstruction for Decoding-Time Controllable Text Generation
This addresses a key limitation in controllable text generation for users needing high-fluency outputs, though it appears incremental as it builds on existing decoding-time methods.
The paper tackles the problem of Attribute Collapse in decoding-time controllable text generation, where text fluency drops sharply at high control strengths, and proposes Air-Decoding, a lightweight framework that reconstructs attribute distributions to balance attribute and non-attribute words, achieving new state-of-the-art control performance.
Controllable text generation (CTG) aims to generate text with desired attributes, and decoding-time-based methods have shown promising performance on this task. However, in this paper, we identify the phenomenon of Attribute Collapse for the first time. It causes the fluency of generated text to rapidly decrease when the control strength exceeds a critical value, rendering the text completely unusable. This limitation hinders the effectiveness of decoding methods in achieving high levels of controllability. To address this problem, we propose a novel lightweight decoding framework named Air-Decoding. Its main idea is reconstructing the attribute distributions to balance the weights between attribute words and non-attribute words to generate more fluent text. Specifically, we train prefixes by prefix-tuning to obtain attribute distributions. Then we design a novel attribute distribution reconstruction method to balance the obtained distributions and use the reconstructed distributions to guide language models for generation, effectively avoiding the issue of Attribute Collapse. Experiments on multiple CTG tasks prove that our method achieves a new state-of-the-art control performance.