Machine Unlearning on Tiny-ImageNet Forget 50% benchmark leaderboard · SOTA2 Research

Machine Unlearning on Tiny-ImageNet Forget 50%

96.6RMIA AUCSCRUB

Evaluation Results

Method	Links
SCRUB2026.04		96.6	—	2.6	60.4	49.1	0	0
FT2026.04		95.5	—	1.1	67.4	47.5	0	0
IEU w/GA+Noisy2026.04		95.5	—	0.4	66.7	45.1	0	0
IEU w/GA2026.04		95.3	—	0.5	66.2	45.6	0	0
IEU w/Noisy2026.04		95.1	—	0.5	66.3	45.9	0	0
SALUN2026.04		94.3	—	6.8	39.7	38.1	0.1	0.1
Retraining2026.04		94	—	0	66.5	44.7	0	0
RL2026.04		93.9	—	2.2	60.3	49.5	0	0
Retraining2026.04		90.9	—	0	94.8	60.8	0	0
FT2026.04		90.9	—	1.3	90.2	62.7	0	0
RL2026.04		90.9	—	2.1	85.8	62.3	0	0
SALUN2026.04		90.9	—	7.7	63.1	55.1	0.4	0.4
IEU w/GA2026.04		90.9	—	1.3	90.5	62.9	0	0
IEU w/GA+Noisy2026.04		90.9	—	1.2	89.6	61.5	0	0
SCRUB2026.04		90.8	—	3.3	81.4	63.8	0	0
IEU w/Noisy2026.04		90.8	—	0.9	90.7	60.5	0	0
NEGGRAD2026.03		79.84	61.19	23.61	—	—	—	—
BASE2026.03		79.74	61.2	23.58	—	—	—	—
SSD2026.03		74.84	38.1	21.09	—	—	—	—
FINETUNE2026.03		61.46	45.74	5.7	—	—	—	—
AMUN2026.03		59.37	48.21	5.34	—	—	—	—
SALUN2026.03		58.04	47.77	5.36	—	—	—	—
NEGGRAD+2026.03		56.66	43.63	4.12	—	—	—	—
ℓ1-SPARSE2026.03		55.23	43.48	4.05	—	—	—	—
RETRAIN2026.03		50.3	47.95	0	—	—	—	—
REGUN2026.03		47.88	45.57	1.37	—	—	—	—