Visual Question Answering on MLLMU-Bench Forget 1.0
29.8AccuracyVisual Knowledge Distillation
Evaluation results
| Method | Links | |
|---|---|---|
| Visual Knowledge DistillationBackbone=LLaVA-1.5-7B2025.12 | 29.8 | |
| MMUBackbone=LLaVA-1.5-7B2025.12 | 31.2 | |
| GA DiffBackbone=LLaVA-1.5-7B2025.12 | 40 | |
| MANUBackbone=LLaVA-1.5-7B2025.12 | 41.6 | |
| KL MinBackbone=LLaVA-1.5-7B2025.12 | 42.4 | |
| Visual Knowledge DistillationBackbone=Qwen2-VL-7B2025.12 | 42.8 | |
| GABackbone=LLaVA-1.5-7B2025.12 | 43.2 | |
| NPOBackbone=LLaVA-1.5-7B2025.12 | 43.2 | |
| MANUBackbone=Qwen2-VL-7B2025.12 | 43.2 | |
| MMUBackbone=Qwen2-VL-7B2025.12 | 44 | |
| KL MinBackbone=Qwen2-VL-7B2025.12 | 45.6 | |
| VanillaBackbone=LLaVA-1.5-7B2025.12 | 45.8 | |
| NPOBackbone=Qwen2-VL-7B2025.12 | 49.6 | |
| GABackbone=Qwen2-VL-7B2025.12 | 50.4 | |
| GA DiffBackbone=Qwen2-VL-7B2025.12 | 54.4 | |
| VanillaBackbone=Qwen2-VL-7B2025.12 | 55.2 |