Unlearning

Benchmarks

Dataset Name	SOTA Method	Metric
MUSE-News 1.0 (test)	F target	Privacy Leak-99.8114	55	19d ago
Syntax-preserving dataset (forget set)	GA	PrivLeak0.51	40	1mo ago
MUSE-Books 1.0 (test)	PRISM	Unlearn Score86	24	19d ago
DomainNet c → s	Original (SF(DA)2)	A_Dr66.6	22	8d ago
DomainNet Sketch to Painting	Original (SF(DA)2)	A_Dr71.3	22	8d ago
HumanEval State unlearning	NL-based	Unlearn Efficacy100	18	16d ago
TOFU Neighbor Set	DirectQA	FB Score66.85	17	3d ago
TOFU Forget Set	ELUDe	FB68.28	17	3d ago
MUSE-Books Harry Potter forget set 500 samples		R-Forget-50039.99	13	1mo ago
MUSE-Books Harry Potter 100 samples (forget set)		R-Forget32.13	13	1mo ago
DomainNet Average	SHOT	A_Dr67.8	11	8d ago
DomainNet r → p	Original (SF(DA)2)	A_Dr75	11	8d ago
DomainNet r → c	Unlearned(+)	A_Dr65.7	11	8d ago
DomainNet r → s	Original (SF(DA)2)	A_Dr55.9	11	8d ago
DomainNet p → r	SHOT	A_Dr78.5	11	8d ago
ToFU		Probability0.9908	10	1mo ago
TOFU (forget01)	OBLIVIATE	Forgetting Quality (F.Q.)92	10	1mo ago
WMDP bio	Task Vector	WMDPbio Score0.628	9	19d ago
WMDP	Task Vector	WMDP Score0.489	9	19d ago
CIFAR-10 Random Forget (50%)		Unlearn Acc100	9	1mo ago
CIFAR-10 Random Forget (10%)	SalUn	Unlearn Accuracy99.25	9	1mo ago
CIFAR-10 50% class-wise forgetting (train test)		Utility Retention (Train, D^r)99.9	8	5d ago
CIFAR-10 30% class-wise forgetting (train test)		Utility Retention D^train_r99.3	8	5d ago
OfficeHome T1 1.0 (test)	Original (SF(DA)2)	A -> C63.6	6	8d ago
Noise-based attacks ten celebs	AP	GPT39.63	6	29d ago

Showing 25 of 32 rows