different-category decoys
what's hiding on my shelf?
fairness: different categories; the frog is fully visible.
stage 1.2 difficulty preview
the landing demo stays easy. this page pressure-tests the real quiz default: tighter crops, more plausible decoys, and a fairness note for every question.
the rule
if a present person cannot point to concrete evidence in the crop, the question gets rejected. the latte foam case below is included as the boundary we do not cross.
calibration set
same answer, different crop and decoy pressure.
fairness: different categories; the frog is fully visible.
fairness: two ceramic shelf objects are plausible; the frog's leg and eye show.
fairness: all options are small glazed ceramic shelf objects; the eye and webbed-foot edge decide it.
calibration set
the reject case is the line: hard cannot mean unknowable.
fairness: color and leaf art are obvious in the crop.
fairness: all options are espresso-milk drinks; foam depth and cup size are visible enough to decide.
fairness: fails: latte vs flat white from foam texture alone is essentially unknowable, even to a barista.
calibration set
Mia's six images are too distinct to produce a fair hard source-guess.
fairness: the sources are wildly distinct; the wristband clearly belongs to the show.
fairness: the dim crop adds work, but the wristband and stage color still point to the show.
calibration set
a hard-but-fair detail from another Mia image.
fairness: the knit weave and mustard color identify the blanket — towel, corduroy, and fleece each show a different, non-knit surface in the crop.
intentional gap
hard source-guess is intentionally missing here: Mia's images are too distinct. Real camera rolls create the confusable source sets this fixture cannot.