Assessing the consistency of the CPL dark-energy model with cosmological data: Simulation-calibrated neural posterior disagreement
PreprintarXiv:2610.310442026
First public:
- Cosmology & gravitation
- AI for science
Research summary
A simulation-calibrated comparison of neural posterior estimators tests the adequacy of the CPL dark-energy model. The combined CMB, DESI and DESY5 observations are consistent with the model under this test, which evaluates rapidly after training and calibration for repeated model-adequacy assessments.
Extended research summary
This study checks whether current cosmological observations are compatible with the CPL description of evolving dark energy. It compares two neural posterior estimators and calibrates their disagreement with independent simulations. Applied to combined CMB, DESI and DESY5 data, the test returns a global p-value of 0.524, finding no inconsistency with CPL in this analysis.
Extended notes
Calibration uses independent simulated data sets to place posterior disagreement on a global statistical scale. The observed-data test combines compressed CMB, DESI and DESY5 measurements.