Machine Unlearning on CIFAR-10 Random Forget 10% (train) benchmark leaderboard · SOTA2 Research

Machine Unlearning on CIFAR-10 Random Forget 10% (train)

100Retain AccuracyRetrain

Evaluation Results

Method	Links
Retrain2025.12		100	94.49	94.33	12.53	0	—	—
RL2025.12		100	100	94.45	3.06	3.77	—	—
BS2025.12		100	100	94.48	1.11	4.27	—	—
SalUn2025.12		100	100	94.47	2.39	3.95	—	—
GD+Sine2025.09		99.98	—	98.59	3.08	0.1	98.71	—
RT2025.05		99.98	10.54	89.59	18.41	0	—	105.7
GA2025.05		99.97	0.06	90.89	0.98	7.31	—	0.38
RT2025.05		99.96	10	89.81	15.14	0	—	17.97
Retrain2025.09		99.94	—	98.85	2.94	0	99.12	—
FT2025.05		99.94	0.25	90.31	1.31	6.03	—	1.84
GD+Tanh2025.09		99.93	—	98.4	3.25	0.25	98.55	—
FT2025.05		99.93	0.42	90.99	3.71	6.57	—	5.31
GA2025.05		99.93	0.24	87.67	4.2	5.72	—	0.17
RL2025.05		99.9	14.98	88.53	57.5	12.17	—	7.38
LetheViT2025.09		99.88	—	98.48	3.21	0.34	98.62	—
SalUn2025.09		99.82	—	98.31	3.61	0.54	98.45	—
NOVO2025.09		99.78	—	98.25	3.47	0.59	98.38	—
NegGrad+2025.05		99.77	2.63	89.88	4.57	4.55	—	2.52
FT2025.09		99.71	—	98.02	4.15	0.83	98.24	—
RL2025.05		99.69	4.14	90.16	21.93	2.7	—	6.21
GD+LoRA2025.09		99.62	—	97.91	6.24	1.55	97.82	—
MCUβ2025.05		99.58	9.99	88.31	14.55	0.62	—	5.88
Fast-NTK2025.09		99.45	—	97.72	5.28	1.22	97.89	—
SalUn2025.05		99.34	11.3	89.66	22.5	2.36	—	5.96
MCU2025.05		99.14	9.71	88.22	14.73	0.77	—	5.89
NegTV2025.05		99.08	2.36	88.53	4.14	6.1	—	0.7
MCU2025.05		98.97	9.52	89	16.33	1.18	—	6.8
GA2025.09		98.85	—	96.48	8.72	2.37	96.31	—
MCUβ2025.05		98.69	10.29	89.11	16.45	1	—	6.82
NegGrad+2025.05		98.63	7.03	89.26	11.71	2.97	—	2.96
NegTV2025.05		98.06	4.58	85	5.23	5.51	—	0.32
SalUn2025.05		97.87	6.67	90.54	35.45	5.99	—	6.38
Amun+Salun2025.12		97.71	94.19	91.79	11.66	1.51	—	—
Amun2025.12		97.47	94.28	91.67	11.61	1.61	—	—
SFRon2025.05		92.86	14.07	85.74	14.32	4.65	—	2.04
SFRon2025.05		88.46	20.48	84.09	26.28	9.71	—	6.23
GA2025.12		5.07	4.77	5.09	32.63	76.58	—	—