A February 7, 2026 publication reports using AlphaEvolve to search for activation functions, as part of broader research into discovering AI architectures.
A publication dated February 7, 2026 says DeepMind is using AlphaEvolve to discover activation functions. It frames this work as part of broader research aimed at accelerating the discovery of new model architectures.
The publication notes that research on activation functions may contribute to new architectures for models and self-improving AI systems. To check the scope and specific findings, consult the original publication and verify the methods and evidence it presents; the available summary gives no detailed experimental results.