Over-Prudence Evaluation on VLGuard
4.48RR (Before)Mixed-SFT
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Mixed-SFTBase Model=LLaVA-1.5-7B, Fine-tuning Strategy=Full fine-tuning2025.03 | 4.48 | 91.76 | |
| Posthoc-SFTBase Model=LLaVA-1.5-7B, Fine-tuning Strategy=Full fine-tuning2025.03 | 2.69 | 90.83 | |
| NPO-UnlearningBase Model=LLaVA-1.5-7B, Fine-tuning Strategy=Full fine-tuning2025.03 | 2.51 | 11.69 | |
| RMU-UnlearningBase Model=LLaVA-1.5-7B, Fine-tuning Strategy=Full fine-tuning2025.03 | 1.25 | 7.56 | |
| LLaVA-1.5-7BFine-tuning Strategy=Full fine-tuning2025.03 | 0.36 | 0.36 | |
| Unsafe-FilterBase Model=LLaVA-1.5-7B, Fine-tuning Strategy=LoRA fine-tuning2025.03 | 0.36 | 0 |