Inha University announced on June 17 that the research team led by Professor Seo Youngdeok from the Department of Electrical and Computer Engineering has developed a new personalization technology for generative artificial intelligence (AI) that can reflect user preferences and tastes without additional training or personalized models.


The research team proposed a method to more precisely capture user preferences and interests in generative AI through the AI model ‘FAN’ (Foundation Encoders Are All You Need).


FAN identifies user preference information for a current query based on previously entered commands (prompts), as well as selection and rating histories, and incorporates these preferences into the generation process by leveraging the structure of existing foundation models.


Because it utilizes the representational capabilities of already-trained ‘foundation encoders’ without requiring personalization modules or additional learning, it enables generative AI models to achieve personalization performance while minimizing additional computational costs.


Core architecture of the artificial intelligence model 'FAN'. Provided by Inha University

Core architecture of the artificial intelligence model 'FAN'. Provided by Inha University

View original image

This research demonstrated that high-quality personalized images can be synthesized in various text and image diffusion models. It has drawn attention from academia for its potential to expand into diverse fields, including multimodal search, image-conditioned diffusion models, and vision-language models (VLM).


The research results were presented at the Conference on Computer Vision and Pattern Recognition.



Professor Seo stated, "As generative AI becomes more widespread, the ability to understand users’ unique preferences and contexts is becoming increasingly important. We aim to create an AI environment that anyone can easily utilize by developing generative AI technology that naturally reflects individual tastes without additional training or complex model design."


This content was produced with the assistance of AI translation services.

© The Asia Business Daily. All rights reserved. Unauthorized AI training and use prohibited.

Today’s Briefing