An Invariant Learning Characterization of Controlled Text Generation

Carolina Zheng, Claudia Shi, Keyon Vafa, Amir Feder, David Blei


Abstract
Controlled generation refers to the problem of creating text that contains stylistic or semantic attributes of interest. Many approaches reduce this problem to training a predictor of the desired attribute. For example, researchers hoping to deploy a large language model to produce non-toxic content may use a toxicity classifier to filter generated text. In practice, the generated text to classify, which is determined by user prompts, may come from a wide range of distributions. In this paper, we show that the performance of controlled generation may be poor if the distributions of text in response to user prompts differ from the distribution the predictor was trained on. To address this problem, we cast controlled generation under distribution shift as an invariant learning problem: the most effective predictor should be invariant across multiple text environments. We then discuss a natural solution that arises from this characterization and propose heuristics for selecting natural environments. We study this characterization and the proposed method empirically using both synthetic and real data. Experiments demonstrate both the challenge of distribution shift in controlled generation and the potential of invariance methods in this setting.
Anthology ID:
2023.acl-long.179
Volume:
Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
Month:
July
Year:
2023
Address:
Toronto, Canada
Editors:
Anna Rogers, Jordan Boyd-Graber, Naoaki Okazaki
Venue:
ACL
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
3186–3206
Language:
URL:
https://aclanthology.org/2023.acl-long.179
DOI:
10.18653/v1/2023.acl-long.179
Bibkey:
Cite (ACL):
Carolina Zheng, Claudia Shi, Keyon Vafa, Amir Feder, and David Blei. 2023. An Invariant Learning Characterization of Controlled Text Generation. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 3186–3206, Toronto, Canada. Association for Computational Linguistics.
Cite (Informal):
An Invariant Learning Characterization of Controlled Text Generation (Zheng et al., ACL 2023)
Copy Citation:
PDF:
https://preview.aclanthology.org/naacl24-info/2023.acl-long.179.pdf
Video:
 https://preview.aclanthology.org/naacl24-info/2023.acl-long.179.mp4