Auditors ask which data produced a model. What records this?
Why A is the answer
ML Metadata lineage
ML Metadata records artifacts and their lineage, so it can show which data produced a model.
In detail
- ML Metadata records lineage, including which data produced a model.
- Lineage connects artifacts and runs.
Use case
An illustrative scenario, not a real incident.
An auditor asks which data trained a model, and lineage in ML Metadata shows it.
See it in different ways
flowchart LR A["Run with parameters"] --> B["Experiments logs metrics"] B --> C["Compare runs"] C --> D["ML Metadata records lineage"]
mindmap
root((ML data))
Processing
Dataflow
Features
Feature Store
Prototyping
Workbench
Colab Enterprise
Tracking
Experiments
ML Metadata lineageflowchart LR S0["Model"] --> S1["Training run"] --> S2["Dataset version"]
Key point: ML Metadata records lineage.
Why the other options are wrong
- B. A safety filter on the model
- A filter is about safety.
- C. A public endpoint for the model
- An endpoint serves predictions.
- D. Feature attribution for one prediction
- That explains one prediction.