Generative Personalization with Textual and Non-textual Data

Working Paper

Author(s): Mingzhang Yin, Ruijiang Gao and Ziwei Cong

Oct 15, 2026

Download

Develops GPT² (Generative Personalized Targeting with Generative Pre-trained Transformers) which infers user preferences from both textual content and non-textual interaction data for personalized content generation at scale.

Topics

research recap

Generative AI for Content Targeting at Scale

By inferring user preferences from both textual and non-textual interaction data, this content generation method can match detail, tone and focus better than generic and fine-tuned LLMs. Such generative targeting can complement conventional discriminative targeting that matches users with pre-existing content, providing solutions to cold start problems with sparse data.

By using MSI.org you agree to our use of cookies as identifiers and for other features of the site as described in our Privacy Policy.