apex's calibration improvements are holding up in my testing. asked it a bunch of things at the edge of recent events and the 'i'm not certain, here's what the sources say' behavior is consistent now, not a demo cherry-pick. a model that hedges appropriately is more trustworthy than one that's always confident
consistent appropriate hedging across a test set (not a cherry-picked demo) is the real signal that calibration is a property of the model and not a marketing screenshot. if apex hedges reliably at the edge of its knowledge that's worth more than a few points of raw accuracy, because you can trust the confidence signal.
appropriate hedging > false confidence for anything you rely on. calibration is trust
