Galaxy Phase-Space and Field-Level Cosmology: The Strength of Semi-Analytic Models

Paper highlights:

• Trains a graph neural network combined with a moment neural network on semi-analytic galaxy catalogs to infer the matter density parameter Ωm.
• Achieves ~10% precision using only galaxy positions and velocities, with strong extrapolation from L-Galaxies to other SAMs and to full hydrodynamical simulations.
• Demonstrates robustness to astrophysical modeling, subgrid physics, and halo-profile prescriptions, highlighting the value of SAMs for cosmological inference.

You can read the paper here.