Document Text Generation on OHR-Bench Generation
49.4Text ScoreGround Truth
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Ground TruthCategory=Ground Truth2025.05 | 49.4 | 46 | 34 | 47 | 28.2 | 43.9 | |
| DiT-Qwen2.5-VL-72BCategory=VLM for OCR, Layout Analysis=DiT2025.05 | 48.7 | 41.8 | 29.7 | 26.6 | 24.3 | 39.8 | |
| SCAN-Qwen2.5-VL-72BCategory=VLM for OCR, Layout Analysis=SCAN2025.05 | 48.4 | 43.3 | 31.9 | 27.6 | 26.7 | 40.8 | |
| Beehive-Qwen2.5-VL-72BCategory=VLM for OCR, Layout Analysis=Beehive2025.05 | 46.6 | 11.3 | 21.3 | 10.2 | 27.2 | 28.2 | |
| DiT-GOT-580MCategory=OCR-specialized small VLM, Layout Analysis=DiT2025.05 | 46.4 | 35 | 25.5 | 19.6 | 14.8 | 35 | |
| SCAN-GOT-580MCategory=OCR-specialized small VLM, Layout Analysis=SCAN2025.05 | 46.2 | 37.9 | 27.5 | 20.8 | 24.6 | 36.9 | |
| Beehive-GOT-580MCategory=OCR-specialized small VLM, Layout Analysis=Beehive2025.05 | 45.7 | 33.3 | 26.4 | 23.7 | 28.1 | 35.9 | |
| olmOCR-7BCategory=VLM for OCR2025.05 | 44.8 | 40.5 | 30.4 | 19 | 8.4 | 36 | |
| Qwen2.5-VL-72BCategory=VLM for OCR2025.05 | 44.3 | 42.1 | 31.8 | 27 | 11.6 | 37.5 | |
| DocLayout-YOLO-GOT-580MCategory=OCR-specialized small VLM, Layout Analysis=DocLayout-YOLO2025.05 | 41.9 | 31.4 | 24.5 | 22.2 | 18.3 | 32.7 | |
| InternVL2.5-78BCategory=VLM for OCR2025.05 | 41.7 | 41.8 | 29 | 33.6 | 3.3 | 35.8 | |
| DocLayout-YOLO-Qwen2.5-VL-72BCategory=VLM for OCR, Layout Analysis=DocLayout-YOLO2025.05 | 41.2 | 10.2 | 19 | 7.2 | 18.2 | 24.4 | |
| GOT-580MCategory=OCR-specialized small VLM2025.05 | 37.5 | 28.5 | 24.1 | 8.5 | 7.1 | 27.8 | |
| Nougat-350MCategory=OCR-specialized small VLM2025.05 | 36.7 | 22.9 | 22.9 | 6.4 | 6.9 | 25.5 |