MUSE
Benchmarks
| Task Name | Dataset Name | SOTA Result | Trend | |
|---|---|---|---|---|
| Machine Unlearning | MUSE Books | Privacy Leakage-89.6 | 83 | |
| Machine Unlearning | MUSE NEWS | VerbMem (Df)58.42 | 34 | |
| Machine Unlearning | MUSE-News Llama 2 7B | Privacy Leakage-99.8951 | 27 | |
| Unlearning | MUSE-Books 1.0 (test) | Unlearn Score86 | 24 | |
| Reasoning Segmentation | MUSE (val) | gIoU (overall)48 | 21 | |
| Machine Unlearning | MUSE forget-set | Performance (BF16)0.365 | 16 | |
| Structural Erasure | MUSE | CAD1.979 | 16 | |
| Machine Unlearning | MUSE | VerbMem on DF0 | 16 | |
| Reasoning Segmentation | MUSE (test) | gIoU (overall)42.3 | 16 | |
| Machine Unlearning | MUSE Books | VerbMem (No Q)99.8 | 15 | |
| Machine Unlearning | MUSE-Books Relearn 50% | Forgetting Score (No VerbMem)90.974 | 15 | |
| Machine Unlearning | MUSE (forget set (Df) and retain set (Dr)) | VerbMem (Df)58.4 | 15 | |
| Knowledge Retention | MUSE utility | Util Score18.4 | 14 | |
| Multi-target reasoning segmentation | MUSE (val) | Overall gIoU56.9 | 13 | |
| Unlearning | MUSE-Books Harry Potter 100 samples (forget set) | R-Forget32.13 | 13 | |
| Machine Unlearning | MUSE News | Extraction Strength0.975 | 12 | |
| Machine Unlearning | MUSE Books | FKM (Forget Set Knowledge Memorization)0 | 9 | |
| Machine Unlearning | MUSE News | Rel Score8.3 | 9 | |
| Machine Unlearning | MUSE Books | Rel7.55 | 9 | |
| Knowledge Retention | MUSE Retain set (Dr) | KnowMem56 | 9 | |
| Machine Unlearning | MUSE Books | VerbMem99.56 | 8 | |
| Machine Unlearning | MUSE Books | VerbMem99.8 | 8 | |
| Knowledge Unlearning | MUSE (forget set Df) | VerbMem Df Pre57.9 | 8 | |
| Relearning Attack | MUSE | RAP43 | 8 | |
| Multi-object referring segmentation | MUSE (test) | gIoU53.2 | 7 |
Showing 25 of 31 rows