Tag
The paper introduces GTR, a softmax-free recurrent vision backbone using gated linear attention that achieves 58.9 box AP on COCO with ~1.9 ms latency, distilled from a DINOv3 detection teacher and efficient for high-resolution dense prediction and edge deployment.
Robbyant, an embodied AI company under Ant Group, released LingBot-Vision, a self-supervised vision backbone family ranging from 21M to 1.1B parameters, under Apache-2.0. It matches or beats DINOv3 on several depth and segmentation benchmarks despite using less than one third of the training data, highlighting a push for open perception models.