LABS EVALUATION · EXPERIMENTAL
Audit the rule. Do not invent a score.
This workspace describes what the public prediction rule consumed, excluded and emitted. It is a reproducibility and coverage audit, not an accuracy benchmark or scientific validation.
labs-prediction-rule-v1
Non-explicit pair + at least two shared descriptors + at least two additional declared signal matches.
59 / 200 smells appear
141 canonical smells do not appear in the current candidate set.
No performance score
No labeled scientific ground-truth benchmark is published here, so SmellBook does not report accuracy, precision, recall or model quality.
Signal count is descriptive, not confidence.
Candidate signal counts and ordering are not confidence or probability.
strawberry ↔ mango
shared-descriptors · same-category · same-temperature · shared-feelings · shared-context-language · same-sound-label
caramel ↔ baked-apple
shared-descriptors · same-category · same-temperature · shared-feelings · shared-context-language
banana ↔ brown-sugar
shared-descriptors · same-category · same-temperature · shared-feelings · shared-context-language
cut-grass ↔ basil
shared-descriptors · same-category · same-temperature · shared-feelings · shared-context-language
black-pepper ↔ toasted-almonds
shared-descriptors · same-category · same-temperature · shared-feelings · shared-context-language
taxi-interior ↔ bus-interior
shared-descriptors · same-category · same-temperature · shared-feelings · same-texture-label
cinnamon ↔ caramel
shared-descriptors · same-category · same-temperature · shared-feelings · shared-context-language
clove ↔ caramel
shared-descriptors · same-category · same-temperature · shared-feelings · shared-context-language
The same public input records, declared rule and stable ordering produce the same public candidate set for this release.