Diagram Understanding on AI2D (Score and Avg. Rel.)
87.7ScoreGemini 2.5 Flash
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Gemini 2.5 FlashSize=-, mode=instruct2026.04 | 87.7 | — | |
| MiniCPM-o 4.5Size=9B, mode=instruct2026.04 | 87.6 | — | |
| Qwen3-VLSize=8B, mode=instruct2026.04 | 85.7 | — | |
| Qwen3-OmniSize=30B-A3B, mode=instruct2026.04 | 85.2 | — | |
| InternVL3.5Size=8B, mode=instruct2026.04 | 84 | — | |
| Full ModelRatio=1.02026.04 | 80.7 | 100 | |
| VisionZipRatio=0.62026.04 | 79.9 | 96.1 | |
| HAWKRatio=0.62026.04 | 79.9 | 99.1 | |
| HAWKRatio=0.82026.04 | 78.1 | 95.8 | |
| VisionZipRatio=0.82026.04 | 77.9 | 89.2 | |
| HAWKRatio=0.92026.04 | 75.3 | 88.5 | |
| DARTRatio=0.62026.04 | 73.2 | 89.5 | |
| VisionZipRatio=0.92026.04 | 71 | 77.5 | |
| DARTRatio=0.82026.04 | 68.3 | 80.4 | |
| DARTRatio=0.92026.04 | 65.2 | 72.2 |