Against a record you choose. Two columns and a delta — never a number on its own, because −8.2 LUFS means nothing until the reference is beside it.
Compare a −6 LUFS master against a −12 LUFS reference and you measure the gain difference, then call it tone. Watch it come out.
A single mark would need weights, and weights are a judgement, not a measurement. These deltas are measured; a score built on top of them would not be.
Three of the four hold inside tolerance. The low end is the one that separates this master from the reference, and it is the one to work on. That sentence is not a summary of a score — it is what the rows below say, counted. Every tolerance that decides a verdict is printed beside it, so you can disagree with the one you disagree with.
This page is an example, not a measurement. The tolerance shown for tone is the one the code uses; the pairs are from a comparison we ran to build this screen. Yours will be your own.