Can machine learning measure galaxy cluster masses better than physics?

Galaxy clusters are key cosmological tools, but their masses are hard to measure reliably from X-ray, SZ, optical, and dynamical data. Machine learning models trained on simulations can capture non-linear effects and projection artifacts that simpler methods overlook, improving mass estimates and characterizing cluster dynamics and mergers. The catch: success depends critically on how well simulations include baryonic physics, and interpretability remains a bottleneck for precision cosmology.