This document summarizes the Plug and Play Language Models (PPLM) approach for controlled text generation. PPLM allows controlling the attributes of text generated by large pretrained language models. It introduces a latent update method using a discriminator to steer generation toward a target attribute while maintaining fluency. The document outlines the PPLM methodology, discusses hyperparameters like alpha values and bag-of-words approaches, and provides examples of attribute-controlled text generation using the PPLM framework.