C-RADIOv4 (Tech Report)

TL;DR

C-RADIOv4 introduces a unified multi-teacher distillation approach that enhances vision models' capabilities and efficiency across various tasks and resolutions, building on previous C-RADIO models with new teacher sets and features.

Contribution

The paper presents C-RADIOv4, a new vision model family that leverages multi-teacher distillation to improve performance and capabilities while maintaining computational efficiency.

Findings

Significant improvements on downstream tasks.

Enhanced resolution support and efficiency.

Incorporation of new teacher models like SAM3.

Abstract

By leveraging multi-teacher distillation, agglomerative vision backbones provide a unified student model that retains and improves the distinct capabilities of multiple teachers. In this tech report, we describe the most recent release of the C-RADIO family of models, C-RADIOv4, which builds upon AM-RADIO/RADIOv2.5 in design, offering strong improvements on key downstream tasks at the same computational complexity. We release -SO400M (412M params), and -H (631M) model variants, both trained with an updated set of teachers: SigLIP2, DINOv3, and SAM3. In addition to improvements on core metrics and new capabilities from imitating SAM3, the C-RADIOv4 model family further improves any-resolution support, brings back the ViTDet option for drastically enhanced efficiency at high-resolution, and comes with a permissive license.