Explains the types of metadata architecture
Centralised (one repository holding copies), distributed (queries to the sources, no copy), hybrid (a central repository plus live sources) and bi-directional (changes flow back to the sources), with their trade-offs in currency, availability and integration effort.
Can AI do this?
People still lead Few or none of the tested AI models showed they master this. 0 of 4 models in the current benchmark master it.
| Model | Tier | Probability of mastery | Status |
|---|---|---|---|
| gemini-4-argon | Flagship | 46% | Not shown |
| gemini-3.5-flash-lite | Low cost | 47% | Not shown |
| gpt-6-luna | Low cost | 48% | Not shown |
| claude-haiku-5-5 | Low cost | 47% | Not shown |
Mastered means a probability of at least 90% that the model has this knobit, estimated from its graded answers. “Not shown” means the answers do not settle it either way. How the models are tested.
What it looks like in practice
Expected result: Centralised (one repository holding copies), distributed (queries to the sources, no copy), hybrid (a central repository plus live sources) and bi-directional (changes flow back to the sources), with their trade-offs in currency, availability and integration effort.
How it is assessed: Answers a CDMP-style multiple-choice question on the metadata architecture types
All knobits of Metadata management.
Source: DAMA-DMBOK: Data Management Body of Knowledge, 2nd edition (revised), chapter 12: Metadata Management, element MV-8