ResearchBenchmarksClass-Incremental Learning on ImageNet 1000 B0 (10 steps)Class-Incremental Learning on ImageNet 1000 B0 (10 steps)Follow73.58Last Step AccuracyBound20.664834.402448.1461.8776Jul 2, 2026Evaluation resultsMethodMethodLinksLast Step AccuracyAverage AccuracyBoundPar.=11.35Par.=11.352026.0773.58—DRDNPar.=11.47Par.=11.472026.0763.8171.83DyToxPar.=11.36Par.=11.362026.0763.3471.29DERPar.=116.89Par.=116.892026.0760.1668.84DKTPar.=11.81Par.=11.812026.0758.2670.44WAPar.=11.68Par.=11.682026.0755.665.67iCaRLPar.=11.68Par.=11.682026.0722.738.4