Evolution of dark energy reconstructed from the latest observations

We reconstruct evolution of the dark energy (DE) density using a nonparametric Bayesian approach from a combination of latest observational data. We caution against parameterizing DE in terms of its equation of state as it can be singular in modified gravity models, and using it introduces a bias preventing negative effective DE densities. We find a $3.7\sigma$ preference for an evolving effective DE density with interesting features. For example, it oscillates around the $\Lambda$CDM prediction at $z\lesssim0.7$, and could be negative at $z\gtrsim2.3$; dark energy can be pressure-less at multiple redshifts, and a short period of cosmic deceleration is allowed at $0.1 \lesssim z\lesssim 0.2$. We perform the reconstruction for several choices of the prior, as well as a evidence-weighted reconstruction. We find that some of the dynamical features, such as the oscillatory behaviour of the DE density, are supported by the Bayesian evidence, which is a first detection of a dynamical DE with a positive Bayesian evidence. The evidence-weighted reconstruction prefers a dynamical DE at a $(2.5\pm0.06)\sigma$ significance level.