Vision similarity grounding on ImageNet 100 benchmark leaderboard · SOTA2 Research

Vision similarity grounding on ImageNet 100

0.38Accuracy CorrelationMulti-Scale CKA

Evaluation Results

Method	Links
Multi-Scale CKA2026.05		0.38	0.28	100	74
CKA2026.05		0.36	0.3	100	90
DistCorr2026.05		0.31	0.26	100	83
Multi-Scale DistCorr2026.05		0.31	0.25	99	60
SVCCA2026.05		0.29	0.21	55	40
SoftCorr2026.05		0.27	0.45	97	45
RBF-CKA2026.05		0.25	0.17	100	65
Multi-Scale RBF-CKA2026.05		0.25	0.17	97	65
AngShape2026.05		0.21	0.24	100	71
HardCorr2026.05		0.21	0.28	97	46
OrthProc2026.05		0.21	0.24	100	71
LinReg2026.05		0.19	0.21	99	94
IMD2026.05		0.17	-0.11	67	56
AD-CKA2026.05		0.17	0.33	91	66
AD-DistCorr2026.05		0.17	0.21	92	64
AD-RBF-CKA2026.05		0.14	0.06	96	64
RSMDiff2026.05		0.09	-0.41	57	45
RTD2026.05		0.09	-0.18	100	57
ProcDist2026.05		0.08	0.1	100	58
PermProc2026.05		0.07	0.18	72	41
RankSim2026.05		0.07	-0.15	99	71
RSA2026.05		0.06	0.12	100	98
EOS2026.05		0.05	0.09	100	93
GULP2026.05		0.02	0.07	100	92
PWCCA2026.05		-0.02	0.13	99	90
AlignCos2026.05		-0.08	0.08	100	71
2nd-Cos2026.05		-0.08	-0.13	100	78
Jaccard2026.05		-0.11	0.36	100	79
ConcDiff2026.05		-0.11	-0.29	53	43
MagDiff2026.05		-0.16	-0.38	37	37
UnifDiff2026.05		-0.18	-0.34	75	17