Run and result identity
Versioned JSON Schemas, canonical JSON serialization, SHA-256 identifiers, stale-state invalidation, seeds, implementation hashes, and export packages.
Evidence and reproduction
Status is a vector, not a badge
The release reports software behavior, numerical behavior, literature agreement, independent review, experimental comparison, and data pedigree separately. A passing test suite cannot silently promote a model into “validated.”
Current release: internal software and numerical checks passed for the declared cases. No independent scientific or experimental validation. External decisions remain pending until a named reviewer evaluates an immutable artifact for a stated intended use.
Required evidence dimensions
Allowed public states
These state names are controlled vocabulary. Details and basis live in the machine-readable registry.
| Dimension | Allowed states | Question answered |
|---|---|---|
| Software verification | not tested · internal tests passed · independently reproduced | Does code behavior meet declared software requirements? |
| Numerical verification | not assessed · convergence demonstrated · cross-solver agreement demonstrated | How do discretization, solver, and numerical error behave in the declared cases? |
| Literature reproduction | none · partial · reproduced within declared tolerance | Does an immutable result agree with a named publication or table inside a named domain? |
| Independent scientific review | pending · changes requested · approved for declared use | Has an independent reviewer approved the exact artifact and scope? |
| Experimental validation | none · pilot · validated within declared domain | How well does the model represent a controlled real-world referent for its intended use? |
| Data maturity | synthetic · public experimental · evaluated · partner-controlled | What is the pedigree and custody state of model inputs or evaluation data? |
Release snapshot
Evidence states by module
| Module | Software | Numerical | Literature | Independent review | Experiment | Data |
|---|---|---|---|---|---|---|
| FusionSim | internal tests passed | convergence demonstrated | partial | pending | none | evaluated |
| Nuclear Data | internal tests passed | not assessed | partial | pending | none | evaluated |
| Mirror Reactor | internal tests passed | convergence demonstrated | none | pending | none | synthetic |
| Torus Orbit | internal tests passed | convergence demonstrated | none | pending | none | synthetic |
| AlphaTrack | internal tests passed | not assessed | none | pending | none | synthetic |
| DirectConvert | internal tests passed | not assessed | none | pending | none | synthetic |
Reproduction surface
Artifacts shipped with this release
Sources, models, data
Source and model records preserve citations, domains, assumptions, omitted physics, versions, licenses, and unresolved differences. The portfolio view complements the inherited executable registry.
Portfolio model registry · Executable model/source registry · Survey source registry
Portable release metadata
RO-Crate 1.3 metadata, CFF citation metadata, release notes, checksums, and SLSA-compatible provenance statement identify the uploaded artifact.
Verification ≠ validation
Intended use and domain come first
NASA model-and-simulation guidance distinguishes verification against requirements from validation against the real world for an intended use, and requires the validation domain to be recorded. This release applies that boundary consistently.