Ant Group Open-Sources LingBot Vision: 1.1B Params Beat 7B DINOv3
AdinaYakup · x · 2026-07-07
Ant Group has released the LingBot Vision series of self-supervised visual backbone networks under the Apache 2.0 license, with model scales ranging from ViT-S to ViT-g. The core innovation is the "masked boundary modeling" pre-training approach, which enables the model to maintain sharper feature representations at object edges, making it highly suitable for dense spatial perception tasks. Self-reported evaluations show that the 1.1B parameter version outperforms the 7B parameter DINOv3 on the NYU-Depth v2 depth estimation task, demonstrating a significant leap in parameter efficiency.
Related event: Ant Robbyant Open-Sources LingBot Vision Models, Topping Depth Benchmarks(18 posts)→
More from Research
- Alex Townsend posts 200 open problems in numerical linear algebra for humans and AI agents — IgorCarron · 2026-09-11
- Navier-Stokes, Riemann, P vs NP: what this week's math buzzwords mean for you — koltregaskes · 2026-09-11
- Fruit fly brain as an LLM: connectome-driven language model demo goes live — ngxson · 2026-09-11
- Harry Collins: LLMs can't do frontier science because they can't invent new language — whoamisri · 2026-09-11
- The Waymo effect: how AI is quietly making research less collaborative — JohnHammersley · 2026-09-11
- Causal-only attention for non-generative tasks is wasteful, argues HF engineer — antoine_chaffin · 2026-09-11