Online Continual Test-Time Adaptation research benchmarks

Benchmarks

Dataset NameSOTA methodMetricTrendResultsLast Updated

ImageNet-C Severity 5 (test)

CoTTA

63.4Accuracy (Gaussian Noise, ImageNet-C S5)

May 20, 2026

CIFAR-10-C severity 5 (test)

CoDiRe

79.26Gaussian Noise Accuracy (Severity 5)

May 19, 2026

LAION-C κ = 0.5 (severity = 1)

DOCO

66.9Places-L Accuracy

Apr 24, 2026

ImageNet-to-ImageNet-C Average (severity 5)

DOCO

62Accuracy

Apr 24, 2026

NINCO-C severity 5

DOCO

62.1Accuracy (ACC)

Apr 24, 2026

SSB-H.-C severity 5

DOCO

61.9Accuracy

Apr 24, 2026

SUN-C severity 5

DOCO

61.9Accuracy (ACC)

Apr 24, 2026

iNaturalist-C severity 5

DOCO

62.5Accuracy (ACC)

Apr 24, 2026

Texture-C severity 5

DOCO

61.8Accuracy

Apr 24, 2026

Places-C severity 5

DOCO

62.1Accuracy (ACC)

Apr 24, 2026

ImageNet-to-ImageNet-C severity=5, κ=0.1 (test)

DOCO

61.7Places-C Accuracy

Apr 24, 2026

ImageNet-to-ImageNet OOD suite (Places-C, Texture-C, iNatur-C, SUN-C, SSB-H-C, NINCO-C) C (test)

DOCO

62.4ACC (Places-C)

Apr 24, 2026

ImageNet-to-ImageNet-C severity=5, κ=0.3 (test)

DOCO

61.4Accuracy (Places-C)

Apr 24, 2026