There's no single paper about the SPM12 version of the segmentation and how it differs from the old segmentation from my "Unified Segmentation" paper. There were lots of things to fix with the old segmentation, particularly relating to how stuff outside the brain was handled. This extra robustness was introduced into SPM8#s New Segment, but this had the disadvantage that it didn't handle atrophied brains so well. The SPM12 implementation tries to take the best aspects of the default SPM8 Segmentation and its New Segment.
The difference between the old and new segmentations in SPM8 are described in the appendix of Weiskopf, Nikolaus, et al. "Unified segmentation based correction of R1 brain maps for RF transmit field inhomogeneities (UNICORT)." Neuroimage 54.3 (2011): 2116-2124.
Differences between New Segment from SPM8 and the default SPM12 segmentation are described in Malone, Ian B., et al. "Accurate automatic estimation of total intracranial volume: a nuisance variable with less nuisance." Neuroimage 104 (2015): 366-372.
Best regards,
-John