Continual Learning

Learning new tasks without forgetting old ones.

1
Datasets
3
Results
average_accuracy
Canonical metric
Canonical Benchmark

Split CIFAR-100

Canonical class-incremental continual learning benchmark: CIFAR-100 is split into 10 sequential tasks of 10 classes each. Models learn tasks one at a time without access to prior-task data and are evaluated on average accuracy across all tasks after the full sequence.

Primary metric: average_accuracy
View full leaderboard

Top 10

Leading models on Split CIFAR-100.

RankModelaverage_accuracyYearSource
1
SLCA (ViT-B/16)
91.52026paper
2
DualPrompt (ViT-B/16)
86.52026paper
3
L2P (ViT-B/16)
83.92026paper

All datasets

1 dataset tracked for this task.

Related tasks

Other tasks in Methodology.