EM-AI-02 · Subject model · W3
AI model, artifact and training
Model definition, artifact versions, training/fine-tune run, configuration, compute and data provenance. Weights are stored outside the public catalog.
Queued for research
Claude: not-started; Grok: not-started.
Research note, in Russian: Entire research brief pending
Subject boundary and candidate types
- AIModel
- ModelArtifact
- TrainingRun
- TrainingConfiguration
- ComputeAllocation
Deep research questions
- How to distinguish the architecture, a checkpoint and a deploy?
- Which inputs does reproducibility require?
- How to reflect access and use restrictions on training data?
Verifiable invariants
- Training pins input versions
- The same name does not prove the same weights
- A compute grant is not Employment
End-to-end acceptance scenario
Two fine-tunes from one checkpoint, different data licenses and one shared endpoint preserve provenance and restrictions.
Negative case
A new endpoint is treated as a newly trained model.
Approaches to compare
- NIST AI RMF: context, evaluation and risk management
- PROV/SPDX: provenance of data, artifacts and actions
- Model registry/evaluation tooling practice; map system, artifact, run and endpoint
Candidates in the live catalogue
- WM-SFT-004 · ML Model Artifact · 0.3.0-research.1 · installable
Semantic fit requires boundary research; a published model does not by itself complete this card. - WM-AI-007 · AI Model Registry Entry · 0.3.0-research.1 · installable
Semantic fit requires boundary research; a published model does not by itself complete this card. - WM-AI-006 · Model Training / Fine-tuning Run · 0.3.0-research.1 · installable
Semantic fit requires boundary research; a published model does not by itself complete this card.
Result requirements
Every card is executed together with the full research contract: definitions, fields and cardinalities, lifecycle, sources, data mastership, rights, the five object facets, at least eight invariants, positive and negative examples, dependencies, migration and applicability limits.