KAN-based ViTs show slight average incremental accuracy gains over MLP-ViTs in continual learning, but the paper's own data show worse forgetting on CIFAR-100 and worse last-task accuracy on MNIST.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
Exploring Kolmogorov-Arnold Network Expansions in Vision Transformers for Mitigating Catastrophic Forgetting in Continual Learning
KAN-based ViTs show slight average incremental accuracy gains over MLP-ViTs in continual learning, but the paper's own data show worse forgetting on CIFAR-100 and worse last-task accuracy on MNIST.