C-RADIOv4 (Tech Report)

2026-01-24

AI Summary

This work proposes a multi-teacher knowledge distillation framework to construct a stronger general-purpose vision student model without increasing computational complexity. By integrating the strengths of advanced teacher models—SigLIP, DINOv3, and SAM—the approach leverages an Agglomerative Vision Backbone architecture combined with ViTDet design, enabling support for arbitrary input resolutions and improving inference efficiency at high resolutions. The resulting C-RADIOv4 model maintains its parameter count at 412M/631M while significantly enhancing performance on downstream tasks, demonstrating superior generalization capability, efficient high-resolution processing, and more permissive licensing terms.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Large Multimodal Models (LMMs)
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

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.

Problem

• multi-teacher distillation
• unified student model
• any-resolution support
• vision backbone
• downstream tasks

Innovation

• multi-teacher distillation
• any-resolution support
• ViTDet
• vision backbone
• model distillation