This storm has a complicated setup. You have multiple surface lows within a broader upper-level trough, marginal surface temperatures, and large discrepancies in QPF placement between the models.
Temperatures along and east of I-95 are the most uncertain, which are expected to be in the low-to-mid 30s through most of the event. Areas north and west of I-95 will start above freezing, and will cool to at or below freezing depending on how strong the snowfall rates are tonight.
Snowfall rates will be one of the most important factors in determining which areas see 4-8 inches, and which areas get 1-4 inches. The best rates are likely to be along the Maryland/Pennsylvania border into southern New Jersey.
The intensity of the initial coastal low, the secondary coastal low development, the strength of the low over the Great Lakes, and the timing of the strengthening and weakening of these lows are wreaking havoc on the QPF output on the various models.
I tried my best to offer a realistic compromise between the differences in QPF, while also accounting for the surface temperature and snowfall rate issues. I could definitely see a tighter gradient between accumulation amounts than what I have on the map.