I believe KAN hype died off due to practical reasons (e.g. FLOPs from implementation) and empirical results, i.e. people reproduced KANs and they found the clai

I believe KAN hype died off due to practical reasons (e.g. FLOPs from implementation) and empirical results, i.e. people reproduced KANs and they found the claims/results made in the original paper were misleading. Here's a paper showing KANs are no better than MLPs, if anything they are typically worse when comparing fairly. https://arxiv.org/pdf/2407.16674 https://arxiv.org/pdf/2407.16674