List your company
ResearchFormerly WizWand

Class-Incremental Learning on CIFAR100 20-step split

83.99Average AccuracyiGSP
12.43831.01449.5968.166May 19, 2026May 25, 2026Jun 1, 2026Jun 8, 2026Jun 14, 2026Jun 21, 2026Jun 28, 2026
Updated 4mo ago

Evaluation results

MethodLinks
iGSP2026.0583.9976.54
MoE-Adapter(++)2026.0583.7276.2
ZSCL2026.0580.3969.58
CLIP zero-shotmode=zero-shot2026.0575.265.74
LwF-VR2026.0574.5463.54
DER2026.0573.9862.55
iCaRL2026.0573.3264.55
DyTox2026.0571.6257.43
BiC2026.0566.4847.02
LwF2026.0560.6440.56
Fine-tune2026.0559.6943.13
UCIR2026.0558.1740.63
Ours-LWMBackbone=Pretrained ResNet-18, Buffer=02026.0656.1943.16
Ours-LWM+ConBackbone=Pretrained ResNet-18, Buffer=02026.0656.1843.1
PODNet2026.0553.9735.02
DERBackbone=ResNet32, Buffer=20002026.0646.2743.62
iCaRLBackbone=ResNet32, Buffer=20002026.0642.1733.1
BiCBackbone=ResNet32, Buffer=20002026.0639.336.17
Fine-tuningBackbone=Pretrained ResNet-18, Buffer=02026.0637.8716.96
LwFBackbone=ResNet32, Buffer=02026.063726.58
PODNetBackbone=Cosine ResNet32, Buffer=20002026.0636.6727.36
FeTrILBackbone=ResNet32, Buffer=02026.0625.9318.51
SimpleCILBackbone=Cosine ResNet32, Buffer=02026.0615.198.73