Skip to main content
Search

Bo Ni

1 related articles

Subscribe to updates
  1. arXiv

    PEARL generates personalized images from user histories via a reason-reflect loop, improving personalization metrics by 15% on its PMH-IG benchmark

    The work introduces PMH-IG, the first unified benchmark for personalized image generation from realistic user histories (Amazon reviews and Instagram posts), comprising Personalized Scene Generation and Personalized Creative Generation with a multi-axis evaluation protocol, and proposes PEARL, which couples a multimodal reasoner with a frozen image generator in an interleaved reason-reflect loop optimized with differential data reward, achieving an average improvement of 15% across personalization metrics on both tasks.

You’re all caught up