Mini toolsFireRed-OCR
By FireRedTeam
It targets “structural hallucination” in complex documents using progressive training and format-constrained GRPO, and reports strong results on OmniDocBench v1.5 with a Qwen3-VL-based 2B model.
Multimodal
Overview
FireRed-OCR is a framework that specializes large vision-language models into pixel-precise structural document parsing models.
Last updated: July 7, 2026
