
A novel method, Plug and Play with Prompts (PPP), that utilizes prompt tuning to steer the generation of text by large language models in a data and parameter efficient manner.


coremsg

plug-and-play-with-prompts-a-prompt-tuning-approach-for-controlling-text-generation


Plug and Play with Prompts: A Prompt Tuning Approach for Controlling Text Generation


title_rewrite


A low-rank autoregressive reward model can efficiently guide text generation from a base language model while maintaining comparable performance to a more flexible but computationally intensive reward model.


efficient-controlled-text-generation-with-low-rank-autoregressive-reward-models


Efficient Controlled Text Generation with Low-Rank Autoregressive Reward Models
