Incremental Learning research benchmarks

Benchmarks

Dataset NameSOTA methodMetricTrendResultsLast Updated

UCF101

iCaRL + IM

10.54Forgetting Rate

Feb 26, 2026

ImageNet-100 B50, 25 steps (test)

Bound

0.812Avg Accuracy

Feb 26, 2026

ImageNet-100 B50Inc2 (T=25) (test)

DS-AL

75.03Avg Accuracy

Feb 26, 2026

MNIST 5-task incremental setting (test)

D³Former

98.85Avg Accuracy

Feb 26, 2026

ImageNet-1000 10 steps 100 classes per step (test)

Bound

89.27Average Session Accuracy

Feb 26, 2026

CUB 200 2011

F2M

81.07Accuracy (Session 1)

Feb 26, 2026

CIFAR100 T=25

MBN-OOD

69.35Last Accuracy

Mar 12, 2026

CIFAR100 T=20

MBN-OOD

69.81Last Accuracy

Mar 12, 2026

CIFAR100 T=10

MBN-OOD

69.59Last Accuracy

Mar 12, 2026

TGB-Wikipedia

AdvSynGNN

0.954Final Accuracy

Feb 26, 2026

SVHN 5-task incremental setting (test)

D³Former

95.81Accuracy

Feb 26, 2026

CIFAR-100 B100Inc0 (T=0)

SCL-PNC

73.07Accuracy

Feb 26, 2026