A variational mapping approach that reveals and expands future temporal dynamics from folded high-dimensional geometric distance spaces, unfold turns a set of time series into a 4D block of pairwise distances between reframed windows, learns a variational mapper that maps those distances to the next reframed window, and produces horizon-wise predictive functions for each input series. In short: it unfolds the future path of each series from a folded geometric distance representation.