Can wavelet statistics extract cosmology without modeling galaxy bias?

Galaxies cluster in ways that depend on both cosmology and their own physical properties—a degeneracy called tracer bias that corrupts standard measurements. This work uses wavelet scattering transforms to build a statistic robust to bias variation, tested on simulations. The method breaks the Ωm–σ8 degeneracy twice as well as traditional approaches and maintains accuracy across different bias scenarios, potentially eliminating a source of systematic error for next-generation surveys.