| Novelty (N) | How unlike the known set a candidate is in your declared CharacteristicSpace. Compute admissibly (declared DescriptorMapRef + DistanceDefRef; no ad-hoc normalisation). | C.17, C.18 |
| Use‑Value (U / ValueGain) | What the candidate helps you achieve now under your CG‑Frame; tie to acceptance/tests; publish units, scale kind, polarity, ReferencePlane. | C.17, C.18 |
| Constraint‑Fit (C) | Satisfies must‑constraints (Resource/Risk/Ethics); legality via CG‑Spec; unknowns propagate (never coerce to zero). | C.18, G.4 |
| Diversity_P (declared retained set) | Coverage or dispersion of the declared retained set under a named measurement policy; declare ReferencePlane. Its change on adding one candidate is DeltaDiversity_P. | C.17, C.18 |
| E/E‑LOG | Named, versioned explore↔exploit policy; governs when to widen space vs refine candidates; policy‑id is published. | C.19 |
| ReferencePlane | Where a value lives: world (system), concept (definition), episteme (about a claim). Plane‑crossings add CL^plane (penalties to R only); cite policy‑id. | F.9, G.6 |
| Scale Variables (S) | The monotone knobs along which improvement is expected (e.g., parameterisation breadth, data exposure, iteration budget, resolution). Declare S for any generator/selector claimed to scale. | C.18.1 |
| Scale Elasticity (χ) | Qualitative class of improvement when moving along S (e.g., rising, knee, flat in the declared window). Used as a selection lens; numeric laws live in domain contexts. | C.18.1 |
| BLP (Bitter‑Lesson Preference) | A preference supported by a comparable, uncertainty-qualified scale comparison; begin with a cheap scale-claim probe. No scale claim yet or no scale-based preference are valid results. A local generality policy is a separate declared basis. | C.19.1, C.24 |
| Iso‑Scale Parity | Fair comparison across candidates at equalised scale budgets along S; may also include scale‑probes (two points) to test elasticity. | G.9, C.18.1 |