Visual Question Answering on MLLMU-Bench 5% forget (Real)
78.59Classification AccuracyVanilla
Evaluation results
| Method | Links | ||
|---|---|---|---|
| VanillaBackbone=Qwen3-VL-8B, Samples=-2026.05 | 78.59 | 0.43 | |
| ASRUBackbone=Qwen3-VL-8B, Samples=27.30%2026.05 | 74.64 | 0.398 | |
| GABackbone=Qwen3-VL-8B, Samples=0%2026.05 | 73.89 | 0.372 | |
| KL_MinBackbone=Qwen3-VL-8B, Samples=100%2026.05 | 73.5 | 0.369 | |
| GA_diffBackbone=Qwen3-VL-8B, Samples=100%2026.05 | 73.11 | 0.356 | |
| NPOBackbone=Qwen3-VL-8B, Samples=0%2026.05 | 71.67 | 0.356 | |
| MMunlearnerBackbone=Qwen3-VL-8B, Samples=100%2026.05 | 71.54 | 0.359 | |
| VanillaBackbone=Qwen3-VL-4B, Samples=-2026.05 | 63.58 | 0.428 | |
| ASRUBackbone=Qwen3-VL-4B, Samples=27.30%2026.05 | 62.53 | 0.382 | |
| MMunlearnerBackbone=Qwen3-VL-4B, Samples=100%2026.05 | 62.27 | 0.38 | |
| GABackbone=Qwen3-VL-4B, Samples=0%2026.05 | 62.14 | 0.377 | |
| KL_MinBackbone=Qwen3-VL-4B, Samples=100%2026.05 | 61.36 | 0.374 | |
| NPOBackbone=Qwen3-VL-4B, Samples=0%2026.05 | 58.75 | 0.371 | |
| GA_diffBackbone=Qwen3-VL-4B, Samples=100%2026.05 | 56.4 | 0.37 | |
| VanillaBackbone=LLaVA-1.5-7B, Samples=-2026.05 | 47.39 | 0.22 | |
| MMunlearnerBackbone=LLaVA-1.5-7B, Samples=100%2026.05 | 44.13 | 0.227 | |
| ASRUBackbone=LLaVA-1.5-7B, Samples=27.30%2026.05 | 43.63 | 0.268 | |
| NPOBackbone=LLaVA-1.5-7B, Samples=0%2026.05 | 43.23 | 0.23 | |
| GABackbone=LLaVA-1.5-7B, Samples=0%2026.05 | 42.27 | 0.23 | |
| KL_MinBackbone=LLaVA-1.5-7B, Samples=100%2026.05 | 41.78 | 0.224 | |
| GA_diffBackbone=LLaVA-1.5-7B, Samples=100%2026.05 | 37.21 | 0.183 |