A Layered Self-Supervised Knowledge Distillation Framework for Efficient Multimodal Learning on the Edge
A layered knowledge distillation framework that compresses a multimodal model for edge deployment.
arXiv, 2025. Preprint, not peer reviewed.
Note
Combines self-supervision with staged distillation so that a compressed multimodal model keeps the discriminative behaviour of the larger one. The target is inference at the point of data collection rather than in a data centre.
Reference
A Layered Self-Supervised Knowledge Distillation Framework for Efficient Multimodal Learning on the Edge arXiv, 2025. https://arxiv.org/abs/2506.07055
The publisher record holds the author list.
Details
- TypePreprint
- AreasFoundation models for biology and medicine
More from the index
- PreviousConfidence-calibrated covariate shift correction for few-shot classification in vision-language modelsCVPR Workshops
- NextVLDBench: Evaluating multimodal disinformation with regulatory alignmentInformation Fusion