• H Company’s new Holo2 model takes the lead in UI Localization Two months since releasing our first batch of Holo2 models, H Company is back with our largest UI localization model yet:Holo2-235B-A22B Preview. • This model achieves a new State-of-the-Art (SOTA) record of 78.5% onScreenspot-Proand 79.0% on OSWorld G. • Available onHugging Face, Holo2-235B-A22B Preview is a research release focused on UI element localization. • Agentic Localization High-resolution 4K interfaces are challenging for localization models. • Small UI elements can be difficult to pinpoint on a large display. • With agentic localization, however, Holo2 can iteratively refine its predictions, improving accuracy with each step and unlocking 10-20% relative gains across all Holo2 model sizes.

Article Summaries:

  • H Company has released its largest UI‑localization model to date, Holo2‑235B‑A22B Preview, which sets new state‑of‑the‑art scores on two industry benchmarks. The model achieves 78.5 % accuracy on Screenspot‑Pro and 79.0 % on OSWorld G, surpassing previous records. Holo2 employs an “agentic” localization approach that iteratively refines predictions, boosting performance by 10‑20 % across model sizes. In a single step the model scores 70.6 % on Screenspot‑Pro, rising to 78.5 % after three refinement steps. Development was accelerated using SkyPilot, a unified cloud interface that simplifies multi‑provider training on Kubernetes clusters.

Sources: