RADIOv2.5: Improved Baselines for Agglomerative Vision Foundation Models
This work addresses efficiency and robustness issues in vision foundation models for AI researchers, though it is incremental.
The paper tackled challenges in agglomerative vision foundation models, such as resolution shifts and teacher imbalance, by proposing multi-resolution training and token compression, resulting in improved baselines with released variants at multiple scales.
Agglomerative models have recently emerged as a powerful approach to training vision foundation models, leveraging multi-teacher distillation from existing models such as CLIP, DINO, and SAM. This strategy enables the efficient creation of robust models, combining the strengths of individual teachers while significantly reducing computational and resource demands. In this paper, we thoroughly analyze state-of-the-art agglomerative models, identifying critical challenges including resolution mode shifts, teacher imbalance, idiosyncratic teacher artifacts, and an excessive number of output tokens. To address these issues, we propose several novel solutions: multi-resolution training, mosaic augmentation, and improved balancing of teacher loss functions. Specifically, in the context of Vision Language Models, we introduce a token compression technique to maintain high-resolution information within a fixed token count. We release our top-performing variants at multiple scales (-B, -L, -H, and -g), along with inference code and pretrained weights