Machine Unlearning on CIFAR10 Subclass airplane (test) benchmark leaderboard · SOTA2 Research

Machine Unlearning on CIFAR10 Subclass airplane (test)

0KL Divergence (t)p_theta_r

Evaluation Results

Method	Links
p_theta_r2026.05		0	—	0	95.9	62	100
pθr2026.05		0	—	0	93.7	49.1	100
pθr2026.05		0	—	0	95.7	60.3	100
LDA-2C2026.05		0.04	0.04	0.35	96.3	63.3	4.8
FT2026.05		0.04	0.04	0.36	97.3	77.6	18
LDA-2C2026.05		0.04	0.05	0.4	94.7	47.1	5
LDA2026.05		0.05	0.05	0.4	95.1	53.6	5.1
SCRUB2026.05		0.05	0.06	0.52	96	64.1	7.2
LDA2026.05		0.05	0.07	0.49	93.9	42.6	5.1
DIR-2C2026.05		0.05	0.05	0.5	99.5	99.2	5
DIR-2C2026.05		0.06	0.06	0.44	96.2	66	5.1
GA+FT2026.05		0.06	0.06	0.62	97.9	82.6	19
LDA-Mix2026.05		0.06	0.08	0.63	98.5	88.3	5.4
DIR2026.05		0.11	0.46	0.63	92.8	40.3	5
RL+FT2026.05		0.11	0.48	0.63	92.8	40.3	4.6
SCRUB2026.05		0.11	3.3	1.1	96.2	65.8	6.7
SalUn2026.05		0.13	0.55	0.6	93.2	51.4	5
GA2026.05		0.14	19	0.68	93.4	57	1.9
LDA-Mix2026.05		0.15	0.15	1.5	98.9	92.3	5.3
DIR-2C2026.05		0.16	0.18	1.5	96.8	97.2	18
LDA-Mix2026.05		0.2	0.21	2	97.1	98.6	21
LDA-2C2026.05		0.2	0.21	2	97.3	99.2	21
LDA2026.05		0.21	0.21	2	97	98.9	21
DIR2026.05		0.22	0.29	0.79	92.4	79.8	18
RL+FT2026.05		0.22	0.29	0.81	92.2	79.8	14
FT2026.05		0.26	0.37	2.7	99.2	95	19
SalUn2026.05		0.28	0.33	0.6	86.2	52	15
p_theta2026.05		0.39	—	3.8	99.5	99.2	—
SCRUB2026.05		0.41	0.45	3.6	95.9	92.4	19
GA+FT2026.05		0.46	0.6	5	97.7	99.7	65
DIR2026.05		0.47	0.62	4.9	89.9	2.6	5
pθ2026.05		0.48	—	4.8	99.7	100	—
RL+FT2026.05		0.48	0.63	5	90	3.3	4.4
pθ2026.05		0.51	—	5.3	98	100	—
FT2026.05		0.58	0.73	6.1	97.8	99.7	64
SalUn2026.05		0.77	0.81	8.1	89.6	0.3	4.7
GA+FT2026.05		0.82	0.84	9.2	91.9	19.5	19
GA2026.05		1.3	2.7	2.9	82.9	45.5	7
GA2026.05		11	24,650	28	60.1	0	2.2