Publication

CAMEO: Correspondence-Attention Alignment for Multi-View Diffusion Models

A training technique that supervises attention maps using geometric correspondence, reducing training iterations by half while achieving superior multi-view generation quality.

AUTHORS Minkyung Kwon*, Jinhyeok Choi*, Jiho Park*, Seonghu Jeon, Jinhyuk Jang, Junyoung Seo, Minseop Kwak, Jin-Hwa Kim, Seungryong Kim
MY ROLE Research Contributor
VENUE CVPR 2026
DATE February 2026
LINKS Paper · Project Page · arXiv · PDF · Code

Contents

  1. Abstract
  2. Links

Abstract

We present CAMEO, a method to align correspondence and attention for multi-view diffusion models. By supervising attention maps with geometric correspondence during training, we ensure that attention across different views remains consistent with underlying 3D geometry. This achieves significantly improved multi-view consistency while reducing training iterations by half.