Explains MDM quality metrics
Duplicate, false-merge and false-split rates, completeness and time to resolve, each defined, measured at a stated point and judged against a target (ISO 8000-8; ISO/IEC 25024).
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 | 3% | Not mastered |
| gemini-3.5-flash-lite | Low cost | 3% | Not shown |
| gpt-6-luna | Low cost | 3% | Not mastered |
| claude-haiku-5-5 | Low cost | 4% | Not mastered |
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: Duplicate, false-merge and false-split rates, completeness and time to resolve, each defined, measured at a stated point and judged against a target (ISO 8000-8; ISO/IEC 25024).
How it is assessed: Answers a written question on the MDM metrics correctly
Related knobits
Builds on
Needed for
All 13 knobits of Run the stewardship workflow and measure MDM.
Source: Practice the stage 5 reference and review checklist rest on