Raphael Douady

Decrypt Modality Gap in Multimodal Contrastive Learning: From Convergent Representation to Pair Alignment

Geometry of Long-Tailed Representation Learning: Rebalancing Features for Skewed Distributions