How accidental was inflation?

Abstract Data on the cosmic microwave background (CMB) are discriminating between different models of inflation, disfavoring simple monomial potentials whilst being consistent with models whose predictions resemble those of the Starobinsky R + R 2 cosmological model. However, this model may suffer from theoretical problems, since it requires a large initial field value, threatening the validity of the effective field theory. This is quantified by the Swampland Distance Conjecture, which predicts the appearance of a tower of light states associated with an effective ultra-violet cutoff. This could be lower than the inflation scale for cases with an extended period of inflation, leading to an additional problem of initial conditions. No-scale supergravity models can reproduce the predictions of the Starobinsky model and accommodate the CMB data at the expense of fine-tuning of parameters at the level of 10 -5 . Here, we propose a solution to this problem based on an explicit realisation of the Starobinsky model in string theory, where this `deformation' parameter is calculable and takes a value of order of the one corresponding to the Starobinsky inflaton potential. Within this range, there are parameter values that accommodate more easily the combination of Planck, ACT and DESI BAO data, while also restricting the range of possible inflaton field values, thereby avoiding the swampland problem and predicting that the initial conditions for inflation compatible with the CMB data are generic.