OCR-based Visual Question Answering on DocVQA 2021 (val)
93.7AccuracyQwen2-VL
Evaluation results
| Method | Links | |
|---|---|---|
| Qwen2-VL2025.01 | 93.7 | |
| DeepSeekVL-2-Small2025.01 | 92.3 | |
| SAIL-VL-8B2025.01 | 92.2 | |
| InternVL2.5-MPO2025.01 | 92.1 | |
| InternVL2.52025.01 | 91.8 | |
| InternVL22025.01 | 90.8 | |
| SAIL-VL2025.01 | 89.2 | |
| Qwen2-VL2025.01 | 88.7 | |
| DeepSeekVL-2-Tiny2025.01 | 88.6 | |
| InternVL2.5-MPOTraining Strategy=MPO2025.01 | 87.8 | |
| InternVL2.52025.01 | 87.4 | |
| InternVL22025.01 | 86 | |
| Aquila2025.01 | 85 |