Multiple Choice Visual Question Answering on NegBench COCO (test)
93.19AccuracyInternVL3.0-8B
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| InternVL3.0-8BParameters=8B2026.03 | 93.19 | 1 | 72.59 | 1 | 95.57 | |
| InternVL2.5-8BParameters=8B2026.03 | 91.21 | 2 | 70.38 | 4 | 96.5 | |
| InternVL3.5-8BParameters=8B2026.03 | 84.29 | 3 | 65.83 | 7 | 93.24 | |
| Qwen2.5-VL-7BParameters=7B2026.03 | 81.65 | 4 | 71.09 | 3 | 83.56 | |
| Qwen3.0-VL-8BParameters=8B2026.03 | 77.53 | 5 | 71.34 | 2 | 77.58 | |
| Gemma-3-12BParameters=12B2026.03 | 71.98 | 6 | 67.15 | 6 | 72.01 | |
| Qwen2.0-VL-7BParameters=7B2026.03 | 69.68 | 7 | 67.43 | 5 | 70.76 | |
| LLaVA-1.5-13BParameters=13B2026.03 | 62.06 | 8 | 58.1 | 8 | 62.06 | |
| InternVL2.0-8BParameters=8B2026.03 | 48.78 | 9 | 39.8 | 10 | 84.34 | |
| LLaVA-1.6-13BParameters=13B2026.03 | 38.98 | 10 | 40.04 | 9 | 44.99 | |
| QwenVL-7B-ChatParameters=7B2026.03 | 24.26 | 11 | 23.1 | 11 | 62.45 |