Visual Question Answering on MLLMU-Bench Retain 1.0
56AccuracyVanilla
Evaluation results
| Method | Links | |
|---|---|---|
| VanillaBackbone=Qwen2-VL-7B2025.12 | 56 | |
| Visual Knowledge DistillationBackbone=Qwen2-VL-7B2025.12 | 56 | |
| MMUBackbone=Qwen2-VL-7B2025.12 | 52.6 | |
| GABackbone=Qwen2-VL-7B2025.12 | 51.5 | |
| NPOBackbone=Qwen2-VL-7B2025.12 | 49.5 | |
| VanillaBackbone=LLaVA-1.5-7B2025.12 | 45.2 | |
| GABackbone=LLaVA-1.5-7B2025.12 | 45 | |
| Visual Knowledge DistillationBackbone=LLaVA-1.5-7B2025.12 | 45 | |
| KL MinBackbone=LLaVA-1.5-7B2025.12 | 44.9 | |
| NPOBackbone=LLaVA-1.5-7B2025.12 | 44.9 | |
| GA DiffBackbone=LLaVA-1.5-7B2025.12 | 44.3 | |
| MMUBackbone=LLaVA-1.5-7B2025.12 | 44.2 | |
| MANUBackbone=LLaVA-1.5-7B2025.12 | 43.2 | |
| MANUBackbone=Qwen2-VL-7B2025.12 | 43.2 | |
| GA DiffBackbone=Qwen2-VL-7B2025.12 | 38.8 | |
| KL MinBackbone=Qwen2-VL-7B2025.12 | 35.9 |