Visual Question Answering on CLEAR 1.0 (Retain)
70.9AccuracyVanilla
Evaluation Results
| Method | Links | |
|---|---|---|
| VanillaBackbone=Qwen2-VL-7B2025.12 | 70.9 | |
| Visual Knowledge DistillationBackbone=Qwen2-VL-7B2025.12 | 70.9 | |
| VanillaBackbone=Qwen2-VL-7B2025.12 | 69.2 | |
| Visual Knowledge DistillationBackbone=Qwen2-VL-7B2025.12 | 69 | |
| KL MinBackbone=Qwen2-VL-7B2025.12 | 68.9 | |
| NPOBackbone=Qwen2-VL-7B2025.12 | 68.9 | |
| MMUBackbone=Qwen2-VL-7B2025.12 | 68.9 | |
| KL MinBackbone=Qwen2-VL-7B2025.12 | 68.4 | |
| NPOBackbone=Qwen2-VL-7B2025.12 | 68.3 | |
| MMUBackbone=Qwen2-VL-7B2025.12 | 68.3 | |
| MANUBackbone=Qwen2-VL-7B2025.12 | 66.2 | |
| GA DiffBackbone=Qwen2-VL-7B2025.12 | 66 | |
| GABackbone=Qwen2-VL-7B2025.12 | 65.9 | |
| MANUBackbone=Qwen2-VL-7B2025.12 | 65.2 | |
| GA DiffBackbone=Qwen2-VL-7B2025.12 | 64.4 | |
| GABackbone=Qwen2-VL-7B2025.12 | 62.4 | |
| VanillaBackbone=LLaVA-1.5-7B2025.12 | 54 | |
| VanillaBackbone=LLaVA-1.5-7B2025.12 | 53.7 | |
| Visual Knowledge DistillationBackbone=LLaVA-1.5-7B2025.12 | 52.8 | |
| GABackbone=LLaVA-1.5-7B2025.12 | 51.8 | |
| NPOBackbone=LLaVA-1.5-7B2025.12 | 51.8 | |
| KL MinBackbone=LLaVA-1.5-7B2025.12 | 51.5 | |
| MMUBackbone=LLaVA-1.5-7B2025.12 | 51.2 | |
| MANUBackbone=LLaVA-1.5-7B2025.12 | 49.8 | |
| Visual Knowledge DistillationBackbone=LLaVA-1.5-7B2025.12 | 49.6 | |
| GABackbone=LLaVA-1.5-7B2025.12 | 48.4 | |
| GA DiffBackbone=LLaVA-1.5-7B2025.12 | 47.7 | |
| MMUBackbone=LLaVA-1.5-7B2025.12 | 46.6 | |
| GA DiffBackbone=LLaVA-1.5-7B2025.12 | 43.4 | |
| NPOBackbone=LLaVA-1.5-7B2025.12 | 38.6 | |
| KL MinBackbone=LLaVA-1.5-7B2025.12 | 38.1 | |
| MANUBackbone=LLaVA-1.5-7B2025.12 | 25.8 |