Ant Group Releases LingBot-Vision Self-Supervised Vision Backbone
Simple_Response8041 · reddit · 2026-07-07
Ant Group has released LingBot-Vision, a DINO-family self-supervised vision backbone available in four sizes (ViT-S/B/L/g, 21M to 1.1B) under the Apache-2.0 license. The core innovation is boundary-driven masking: a teacher network predicts object boundaries and forces these tokens into the student network's mask, preventing it from simply copying flat contexts for reconstruction. The entire process requires no labels, text supervision, or external edge detectors. On NYUv2 depth estimation, the 1.1B flagship achieves an optimal 0.296 RMSE (compared to 0.309 for DINOv3-7B), while the 0.3B ViT-L reaches 0.310, rivaling the 7B model with roughly 23x fewer parameters. Its weak point is ImageNet classification, where both the flagship and L sizes trail DINOv3. The repository includes a lightweight loader to output frozen backbones for feature extraction, focusing on dense features like depth, segmentation, and tracking rather than conversational tasks.
Related event: Ant Robbyant Open-Sources LingBot Vision Models, Topping Depth Benchmarks(18 posts)→
More from Research
- Diamandis: AI Cancer Navigator and $100 Genome Sequencing Close the Care Gap — PeterDiamandis · 2026-10-27
- EMNLP Oral: training-free Nexus Sampling cuts KV-cache memory 10x; OpenResearcher tops BrowseComp-Plus at 54.8% — yuz9yuz · 2026-10-27
- ICML 2026 position paper: LLM 'thinking traces' are a harmful anthropomorphism — rao2z · 2026-10-27
- Forking gap follows a power law in context frequency; masking rare n-grams shrinks it — ZimingLiu11 · 2026-10-27
- How forking happens: n-gram tables split logits into non-interacting subspaces — ZimingLiu11 · 2026-10-27
- The best LoRA explainer video yet: rank constraints meet intrinsic dimension — sophiamyang · 2026-10-27