A number of relatively recent sensors do pixel binning, combining the outputs of multiple sensor elements to make one bigger pixel. You could argue that it's just interpolating those pixels and "downsampling", but it seems to produce better images than what that would imply.