LingBot-Vision: masked boundary modeling for self-supervised pretraining (0.296 NYUv2 linear-probe RMSE at 1.1B vs 0.309 for DINOv3-7B, trails on ImageNet); weights in 4 sizes[R]
LingBot-Vision: masked boundary modeling for self-supervised pretraining (0.296 NYUv2 linear-probe RMSE at 1.1B vs 0.309 for DINOv3-7B, trails on ImageNet); weights in 4 sizes[R] — reported by reddit.com, aggregated and ranked by ClawDigest.