Today for AI
HOT RADAR
opensourceHEAT 11.1°

AI CURATED HOT STORY · 10/30/2026

Xiaomi Open-Sources 0.8B OCR Model for Document Parsing and VQA

1 reports archived1 independent sourcesupdated 10/30/2026, 21:15:17
Synthesis & Latest Updates
1 Sources Cross-Validated

Xiaomi has open-sourced Xiaomi-OCR-0, a lightweight document understanding model built on the Qwen3.5-0.8B base. This single-model architecture covers document parsing, key information extraction, and OCR visual question answering, natively supporting Markdown, LaTeX, and JSON outputs. The training process employs a heterogeneous expert consensus mechanism and a 'predict-re-render-compare' closed-loop verification to significantly enhance accuracy in complex tables, formulas, and long-tail scenarios.

LATEST/Xiaomi has open-sourced Xiaomi-OCR-0, a lightweight document understanding model built on the Qwen3.5-0.8B base. This single-model architecture covers document parsing, key information extraction, and OCR visual question answering, natively supporting Markdown, LaTeX, and JSON outputs. The training process employs a heterogeneous expert consensus mechanism and a 'predict-re-render-compare' closed-loop verification to significantly enhance accuracy in complex tables, formulas, and long-tail scenarios.

Hourly heat curve

2 fully observed hours
Fewer than 3 fully observed hours; trend not plotted yet.

Coverage timeline

Total 1 reports · Latest first
  1. IT之家 · 智能时代T2·78 pts

    Xiaomi Open-Sources 0.8B OCR Model for Document Parsing and VQA

    • Built on Qwen3.5-0.8B base, handling document parsing, info extraction, and VQA with a single model
    • Introduces a 'predict-re-render-visual compare' closed-loop verification to solve complex table and formula parsing issues
    • Trained on ~170 million samples using a heterogeneous expert consensus mechanism to generate high-confidence annotations