Abstract · public
Maintaining an engaging online presence is a persistent challenge for digital enterprises, necessitating alignment with the evolving preferences of their audience. We propose a structure that addresses this issue by using large language models to drive a rapid experimentation system (capable of collecting insights on user behavior and generating recommendations for a more personalized user experience) in which site owners specify entities and valid mutations, models generate candidate widgets inside that grammar, and every proposal is statically checked before serving.
Saussy & Lanzone · full paper shared with design partners.
Request to read ↓
Request drafted, your email app should have opened.
Send it and you'll get a personal reply, usually same-day. Nothing opened? Email us directly.