Text Structural Anomaly Perception on Chinese recognition
100PrecisionMonkeyOCR
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| MonkeyOCRlevel=Image-level2026.02 | 100 | 1.3 | 2.5 | |
| TextPecker (InternVL3-8B)level=Box-level2026.02 | 91.2 | 98.8 | 94.9 | |
| TextPecker (Qwen3-VL-8B)level=Box-level2026.02 | 90 | 96.4 | 93.1 | |
| TextPecker (InternVL3-8B)level=Image-level2026.02 | 88.9 | 96.8 | 92.7 | |
| TextPecker (Qwen3-VL-8B)level=Image-level2026.02 | 87.4 | 98.1 | 92.5 | |
| GOT-OCR-2.0level=Image-level2026.02 | 50 | 0.4 | 0.8 | |
| Qwen3-VL-8Blevel=Image-level2026.02 | 40 | 0.8 | 1.7 | |
| Gemini-2.5-prolevel=Box-level2026.02 | 33.3 | 9.9 | 15.2 | |
| Doubao-Seed-1.6-thinklevel=Box-level2026.02 | 33.3 | 8.6 | 13.7 | |
| Doubao-Seed-1.6level=Box-level2026.02 | 31 | 11.5 | 16.8 | |
| Doubao-Seed-1.6level=Image-level2026.02 | 30.6 | 18.2 | 22.8 | |
| PP-OCRv5level=Image-level2026.02 | 30 | 1.3 | 2.4 | |
| GPT-5level=Box-level2026.02 | 28.2 | 16.5 | 20.9 | |
| GPT-5level=Image-level2026.02 | 23.3 | 22 | 22.6 | |
| Doubao-Seed-1.6-thinklevel=Image-level2026.02 | 21.6 | 8.4 | 12.1 | |
| InternVL3-8Blevel=Image-level2026.02 | 19 | 12.8 | 15.3 | |
| InternVL3-8Blevel=Box-level2026.02 | 12.9 | 37 | 19.1 | |
| Gemini-2.5-prolevel=Image-level2026.02 | 7.9 | 4.8 | 5.9 | |
| Qwen3-VL-8Blevel=Box-level2026.02 | 0 | 0 | 0 |