The CAMELS project: Expanding the galaxy formation model space with new ASTRID and 28-parameter TNG and SIMBA suites

Paper highlights:

• Introduces CAMELS-ASTRID, a new suite of 2,124 hydrodynamical simulations extending the CAMELS project with broader coverage of cosmological and feedback parameters.
• Shows that ASTRID features milder AGN feedback and a wider diversity of galaxy populations, enabling machine-learning models with improved extrapolation across different hydrodynamical simulations.
• Demonstrates that training on highly diverse galaxy-formation models is key for robust ML-based cosmological inference, including accurate parameter recovery from high-dimensional simulation sets

You can read the paper here.