OvisOCR2 Tops Document Parsing Leaderboard

OvisOCR2, a 0.8B document parsing VLM fine-tuned from Qwen3.5-0.8B, has topped parsing leaderboards. It aims to convert single-page document images directly into naturally ordered Markdown, covering text, formulas, and tables.

2026-07-15 ~ 2026-07-16 · 2 related posts

Full story(2 episodes)→