If you are interested in CV application of KANs, there are some interesting GitHub repositories to visit:

https://github.com/chenziwenhaoshuai/Vision-KAN - here ViT like models with KAN layers instead of traditional ones used, with cigar and imagenet benchmarks

https://github.com/IvanDrokin/torch-conv-kan
Here convolutional layers and models presented, with benchmarks on mnist/imagenet/cifar. There are still a lot of work and experiments to do, but some preliminary results are shown