Text Structural Anomaly Perception on English recognition
79.5PrecisionTextPecker (InternVL3-8B)
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| TextPecker (InternVL3-8B)level=Image-level2026.02 | 79.5 | 96 | 87 | |
| TextPecker (InternVL3-8B)level=Box-level2026.02 | 78.4 | 96.4 | 86.5 | |
| TextPecker (Qwen3-VL-8B)level=Image-level2026.02 | 77.7 | 96.9 | 86.2 | |
| TextPecker (Qwen3-VL-8B)level=Box-level2026.02 | 71.4 | 93.8 | 81.1 | |
| Qwen3-VL-8Blevel=Box-level2026.02 | 50 | 1.8 | 3.4 | |
| GPT-5level=Box-level2026.02 | 41.9 | 19.3 | 26.5 | |
| Gemini-2.5-prolevel=Box-level2026.02 | 34.2 | 17.9 | 23.5 | |
| Doubao-Seed-1.6level=Box-level2026.02 | 33.3 | 18 | 23.4 | |
| Qwen3-VL-8Blevel=Image-level2026.02 | 28.6 | 1.7 | 3.2 | |
| Doubao-Seed-1.6-thinklevel=Box-level2026.02 | 28 | 9.5 | 14.1 | |
| Doubao-Seed-1.6-thinklevel=Image-level2026.02 | 25.9 | 18.3 | 21.4 | |
| InternVL3-8Blevel=Box-level2026.02 | 21.8 | 44.3 | 29.3 | |
| InternVL3-8Blevel=Image-level2026.02 | 20.6 | 16.5 | 18.3 | |
| GPT-5level=Image-level2026.02 | 19.6 | 15 | 17 | |
| Gemini-2.5-prolevel=Image-level2026.02 | 17.9 | 7.6 | 10.7 | |
| Doubao-Seed-1.6level=Image-level2026.02 | 15.7 | 16.7 | 16.2 | |
| PP-OCRv5level=Image-level2026.02 | 0 | 0 | 0 | |
| GOT-OCR-2.0level=Image-level2026.02 | 0 | 0 | 0 | |
| MonkeyOCRlevel=Image-level2026.02 | 0 | 0 | 0 |