Field-level simulation-based inference with galaxy catalogs: the impact of systematic effects

Paper highlights:

• Extends graph neural network–based field-level cosmological inference to observational conditions, including masking, spatial and velocity uncertaninties, and galaxy selection effects.
• Tests models on thousands of hydrodynamical simulations from the CAMELS project, incorporating the systematics
• Shows that despite degraded performance, the models succeed for over 90% of galaxy catalogs, highlighting strong potential for application to real survey data.

You can read the paper here.