Single-Class Unlearning on Office 31 D -> W benchmark leaderboard · SOTA2 Research

Single-Class Unlearning on Office 31 D -> W

81.9Accuracy ($A_{D_T}^r$)Unlearned(+)

Evaluation Results

Method	Links
Unlearned(+)2026.04		81.9	100	41
Retrain2026.04		78.4	0	78
Original (SF(DA)²)2026.04		76.4	100	38
ZSMU2026.04		76.2	100	38
Ours2026.04		76.2	0	76
Finetune2026.04		76	100	38
PADA2026.04		75.8	100	38
Nabla Tau2026.04		75.3	0	75
SHOT2026.04		75.1	1.6	74
Lipschitz2026.04		74.5	33.3	62
UNSIR2026.04		67.9	0	68