The Year in AI Papers
The year/Independent research

Paper 2509.18174

Baseer: A Vision-Language Model for Arabic Document-to-Markdown OCR

Khalil Hennara, Muhammad Hreden, Mohamed Motasim Hamed, Ahmad Bastati, Zeina Aldallal, Sara Chrouf, Safwan AlModhayan

Published
Sep 2025
Research lab
Independent
Citations
2
GitHub
Not linked

01 In brief

Summary

The paper introduces Baseer, a vision-language model fine-tuned for Arabic document OCR, and Misraj-DocOCR, a benchmark for evaluation.

Baseer is built on Qwen2.5-VL-3B-Instruct and trained on 500,000 image-text pairs (300k synthetic, 200k real-world) using a decoder-only fine-tuning strategy that freezes the vision encoder.

The authors also corrected the KITAB-bench PDF-to-markdown dataset, addressing errors like hallucinated text and missing page numbers.

Experiments show Baseer achieves a WER of 0.25 on Misraj-DocOCR, outperforming open-source and commercial models, and excels in structural metrics (TEDS, MARS) on both benchmarks.

The study highlights the benefits of domain-specific adaptation of general-purpose MLLMs for morphologically rich languages like Arabic, establishing a strong baseline for high-accuracy OCR.

Key findings include that decoder-only fine-tuning outperforms full fine-tuning and LoRA, and a sequence length of 4096 yields optimal performance.

The benchmark and corrected dataset are publicly available.

02 From the paper

Abstract

Arabic document OCR remains a challenging task due to the language's cursive script, diverse fonts, diacritics, and right-to-left orientation. While modern Multimodal Large Language Models (MLLMs) have advanced document understanding for high-resource languages, their performance on Arabic remains limited. In this work, we introduce Baseer, a vision-language model fine-tuned specifically for Arabic document OCR. Leveraging a large-scale dataset combining synthetic and real-world documents, Baseer is trained using a decoder-only fine-tuning strategy to adapt a pre-trained MLLM while preserving general visual features. We also present Misraj-DocOCR, a high-quality, expert-verified benchmark designed for rigorous evaluation of Arabic OCR systems. Our experiments show that Baseer significantly outperforms existing open-source and commercial solutions, achieving a WER of 0.25 and establishing a new state-of-the-art in the domain of Arabic document OCR. Our results highlight the benefits of domain-specific adaptation of general-purpose MLLMs and establish a strong baseline for high-accuracy OCR on morphologically rich languages like Arabic.