Home/Browse/Computer Vision/Document OCR/KITAB-Bench

KITAB-Bench

MBZUAI

8,809 Arabic text samples across 9 domains. Tests Arabic script recognition.

PaperLeaderboard

Benchmark Stats

Models8
Papers8
Metrics1

SOTA History

Coming Soon
Visual timeline of state-of-the-art progression over time will appear here.

Character Error Rate

Levenshtein distance between predicted and ground truth (lower is better)

Lower is better

RankModelCodeScorePaper / Source
1gemini-20-flash

Arabic OCR - Character Error Rate (lower is better). 8,809 samples, 9 domains

-0.13%AlphaXiv
2ain-7b-0.20%AlphaXiv
3gpt-4o-0.31%AlphaXiv
4gpt-4o-mini-0.43%AlphaXiv
5azure-ocr-0.52%AlphaXiv
6tesseract0.54%AlphaXiv
7easyocr0.58%AlphaXiv
8paddleocr0.79%AlphaXiv
Back to Document OCR