Control text generation for language models like GPT2. We use PPLM, an approach that combines a language model with attribute models to control the style.
S. Banerjee, and A. Lavie. Proceedings of the acl workshop on intrinsic and extrinsic evaluation measures for machine translation and/or summarization, page 65--72. (2005)
D. Oniani, and Y. Wang. Proceedings of the 11th ACM International Conference on Bioinformatics, Computational Biology and Health Informatics, page 1--9. (2020)
L. Chen, D. Zhang, and M. Levene. Proceedings of the 36th international ACM SIGIR conference on Research and development in information retrieval, page 973--976. (2013)