Figure 2 (model distributions): Checkerboard samples from AdS + KG (H), AdS + KG (L), AdS, and the baseline FCN
Inspired by the AdS/CFT correspondence, Holographic Generative Flows introduces a physics-based framework for generative flow matching, accepted in Machine Learning: Science and Technology.

Figure 2 (model distributions): Checkerboard samples from AdS + KG (H), AdS + KG (L), AdS, and the baseline FCN. Source paper, page 12.