We compare double-differential normalized production cross sections for top-antitop $+ X$ hadroproduction at NNLO QCD accuracy, as obtained through a customized version of the MATRIX framework interfaced to PineAPPL, with recent data by the ATLAS and CMS collaborations.

We take into account theory uncertainties due to scale variation and we see how predictions vary as a function of parton distribution function (PDF) choice and top-quark pole mass value, considering different state-of-the-art PDF fits with their uncertainties.

Notwithstanding the overall reasonable good agreement, we observe discrepancies at the level of a few $\sigma$'s between data and theoretical predictions in some kinematical regions, which can be alleviated by refitting the top-quark mass value, and/or the PDFs and/or $\alpha_s(M_Z)$, considering the correlations between these three quantities.

In a fit of top-quark mass standalone, we notice that, for all considered PDF + $\alpha_s(M_Z)$ sets used as input, some datasets point towards top-quark pole mass values lower by about $2 \sigma$'s than those emerging from fitting other datasets, suggesting a possible tension between experimental measurements using different decay channels, and/or the need of better estimating uncertainties on the latter.