Multimodal Question Answering on MLLM Pre-trained Evaluation Suite
78.5VQAv2 AccuracyPre-trained Model
Evaluation results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Pre-trained ModelBackbone=LLaVA-1.5-7B, Target Task=OKVQA2026.06 | 78.5 | 61.9 | 50 | 70.4 | 58.2 | 87.3 | 64.3 | 58.3 | 66.1 | |
| CGM†Backbone=LLaVA-1.5-7B, Target Task=OKVQA2026.06 | 76.2 | 61.4 | 49.5 | 70.6 | 55.3 | 87.4 | 65.1 | 59.5 | 65.7 | |
| CGMBackbone=LLaVA-1.5-7B, Target Task=OKVQA2026.06 | 74.3 | 58.5 | 52.3 | 69.3 | 53.8 | 86.2 | 63.7 | 55.6 | 64.2 | |
| GraftingBackbone=LLaVA-1.5-7B, Target Task=OKVQA2026.06 | 72.5 | 56.1 | 44 | 70.6 | 51 | 88.5 | 62 | 51.4 | 62 | |
| TailorBackbone=LLaVA-1.5-7B, Target Task=OKVQA2026.06 | 71.7 | 56.1 | 40.1 | 69.8 | 50.5 | 82.3 | 64.4 | 58.2 | 61.6 | |
| MagnitudeBackbone=LLaVA-1.5-7B, Target Task=OKVQA2026.06 | 69.7 | 52.8 | 41.5 | 33 | 44.6 | 84.5 | 62.7 | 52.8 | 54 | |
| WandaBackbone=LLaVA-1.5-7B, Target Task=OKVQA2026.06 | 68.8 | 51.2 | 39.1 | 30.3 | 41.4 | 83.1 | 58.8 | 45.1 | 52.2 | |
| Fine-tuned ModelBackbone=LLaVA-1.5-7B, Target Task=OKVQA2026.06 | 68 | 50.4 | 38.7 | 24.5 | 40.1 | 83.2 | 56.9 | 41.6 | 50.4 | |
| DAREBackbone=LLaVA-1.5-7B, Target Task=OKVQA2026.06 | 67.7 | 49.6 | 37.2 | 28.6 | 39.9 | 82.4 | 57.2 | 45.2 | 51 |