Meta paper introduces JEM self-supervised framework rivaling DINO
A 12x reduction in training data to beat a frontier SSL checkpoint suggests the data appetite of self-supervised pretraining is not fixed, which matters for teams with limited compute budgets. Researchers doing vision pretraining or fine-tuning DINO-family backbones should review the JEM method and regularization design for reuse in their own pipelines. Compared with DINOv3, it trades scale for a simpler, more principled objective.
Meta researchers published a paper titled "Scalable Patch-Level Self-Supervised Learning" describing a new self-supervised learning framework called JEM. JEM is a student-teacher method that aligns patch representations across different views and uses explicit regularization to keep representations informative and prevent structural collapse. The authors report that JEM outperforms DINOv2 in a controlled setting on dense tasks and beats the DINOv3 checkpoint while training on 12x less data.
This paper went completely under the radar!
Meta researchers have developed a novel and simple self-supervised learning framework competitive with DINOv2 and DINOv3!
"Scalable Patch-Level Self-Supervised Learning"
The authors introduce JEM, a principled student-teacher method that learns to align patch representations across different views. Additionally, it incorporates explicit regularization to ensure representations are informative and prevent structural collapse.
The framework outperforms DINOv2 in a controlled setting on dense tasks, and even beats the DINOv3 checkpoint despite it being trained on 12x less data.
