Image Classification on CIFAR-10 (Epsilon-based Accuracy) benchmark leaderboard · SOTA2 Research

Image Classification on CIFAR-10 (Epsilon-based Accuracy)

90.12Accuracy (ε=0.5)No randomization

Evaluation Results

Method	Links
No randomization2026.05		90.12	90.12	90.12	90.12	90.12	90.12	90.12	90.12	90.12	90.12	90.12	90.12
PrivUnitG + PA2026.05		11.81	13.84	16.12	18.54	21.22	23.99	26.82	29.68	32.55	35.23	47.3	55.91
PrivUnit2 + PA2026.05		11.72	13.75	16.06	18.52	21.15	24	26.92	29.85	32.76	35.65	48.34	57.66
Laplace + PA2026.05		10.64	11.14	11.66	12.21	12.75	13.36	14	14.64	15.32	15.98	19.68	23.8
PrivUnit22026.05		10.55	10.94	11.44	11.91	12.37	12.86	13.36	13.84	14.36	14.92	17.23	19.87
CW + PA w/o bounding2026.05		10.45	10.76	11.09	11.42	11.76	12.11	12.48	12.83	13.21	13.57	15.67	18.01
PrivUnitG2026.05		10.44	10.87	11.33	11.81	12.29	12.78	13.25	13.76	14.24	14.71	17.09	19
Laplace + PR2026.05		10.17	10.19	10.22	10.24	10.27	10.3	10.32	10.35	10.37	10.39	10.52	10.64
Laplace (ℓ1)2026.05		10.15	10.21	10.27	10.33	10.39	10.44	10.5	10.55	10.61	10.68	11	11.34
Laplace (ℓ∞)2026.05		10.11	10.12	10.14	10.15	10.16	10.17	10.19	10.2	10.21	10.23	10.3	10.36