OCR-based Visual Question Answering on AI2D 2016 (test)
84.6AccuracyInternVL2.5
Evaluation results
| Method | Links | |
|---|---|---|
| InternVL2.52025.01 | 84.6 | |
| InternVL2.5-MPO2025.01 | 84.1 | |
| InternVL2.5-MPOModel Size=8B2025.01 | 84.1 | |
| SAIL-VL-8B2025.01 | 83.7 | |
| InternVL22025.01 | 83.7 | |
| SAIL-VLModel Size=8B2025.01 | 83.7 | |
| Qwen2-VL2025.01 | 82.9 | |
| Qwen2-VLModel Size=8B2025.01 | 82.9 | |
| DeepSeekVL-2-Small2025.01 | 82 | |
| DeepSeekVL-2Model Size=8B2025.01 | 82 | |
| SAIL-VL2025.01 | 79 | |
| SAIL-VLModel Size=2B2025.01 | 79 | |
| InternVL2.5-MPOTraining Strategy=MPO2025.01 | 75.3 | |
| InternVL2.5-MPOModel Size=2B2025.01 | 75.3 | |
| InternVL2.52025.01 | 75 | |
| Aquila2025.01 | 75 | |
| DeepSeekVL-2-Tiny2025.01 | 74.6 | |
| DeepSeekVL-2Model Size=2B2025.01 | 74.6 | |
| InternVL22025.01 | 74.2 | |
| Qwen2-VL2025.01 | 72.3 | |
| Qwen2-VLModel Size=2B2025.01 | 72.3 |