Kolmogorov-Arnold Networks with geometry-constrained edge activations match or outperform fixed-basis methods on symbolic regression tasks, achieving median NRMSE of 0.030 on AI Feynman benchmarks

Across 50 symbolic-regression targets ($40$ from the AI Feynman benchmark plus $10$ synthetic stress tests), geometry-constrained KANs match or beat every fixed-basis baseline on median NRMSE (Banach-KAN $0.030$, tying Chebyshev and improving on splines)

http://arxiv.org/abs/2608.25807v1