Visual Question Answering on MLLMU-Bench Real-world 1.0
77.3AccuracyVanilla
Evaluation results
| Method | Links | |
|---|---|---|
| VanillaBackbone=Qwen2-VL-7B2025.12 | 77.3 | |
| Visual Knowledge DistillationBackbone=Qwen2-VL-7B2025.12 | 75.4 | |
| MMUBackbone=Qwen2-VL-7B2025.12 | 75.3 | |
| NPOBackbone=Qwen2-VL-7B2025.12 | 75.2 | |
| KL MinBackbone=Qwen2-VL-7B2025.12 | 74.8 | |
| GA DiffBackbone=Qwen2-VL-7B2025.12 | 74.5 | |
| GABackbone=Qwen2-VL-7B2025.12 | 74.4 | |
| MANUBackbone=Qwen2-VL-7B2025.12 | 73.7 | |
| VanillaBackbone=LLaVA-1.5-7B2025.12 | 47.4 | |
| Visual Knowledge DistillationBackbone=LLaVA-1.5-7B2025.12 | 47.2 | |
| GABackbone=LLaVA-1.5-7B2025.12 | 47 | |
| NPOBackbone=LLaVA-1.5-7B2025.12 | 47 | |
| MMUBackbone=LLaVA-1.5-7B2025.12 | 46.7 | |
| GA DiffBackbone=LLaVA-1.5-7B2025.12 | 46.6 | |
| KL MinBackbone=LLaVA-1.5-7B2025.12 | 46.6 | |
| MANUBackbone=LLaVA-1.5-7B2025.12 | 46.3 |