Robust extractable cosmological information from galaxies and LSS tracers

Determine the amount of cosmological information that can be robustly extracted from galaxies and other large-scale structure tracers, assessing the information content available from the full three-dimensional tracer fields relative to low-order summary statistics such as the power spectrum and bispectrum, in order to quantify the limits of field-level analyses for parameters like the amplitude of linear matter fluctuations σ8.

Background

The paper contrasts field-level Bayesian inference using an EFT-based forward model with simulation-based inference on summary statistics (power spectrum and bispectrum) to assess how much information beyond low-order n-point functions can be extracted from galaxy clustering.

The central motivation is the longstanding reliance on two- and three-point statistics in cosmological analyses and the desire to understand whether more information can be obtained from the full field-level data, particularly for constraining σ8. The authors frame this as an open question and then present empirical results addressing it.