Explains measuring false merges and false splits
False-merge and false-split rates measured on a labelled sample before go-live and periodically; merges counted alone say nothing about quality.
Can AI do this?
AI does this well Most of the tested AI models master this. 3 of 4 models in the current benchmark master it.
| Model | Tier | Probability of mastery | Status |
|---|---|---|---|
| gemini-4-argon | Flagship | 100% | Mastered |
| gemini-3.5-flash-lite | Low cost | 96% | Not shown |
| gpt-6-luna | Low cost | 100% | Mastered |
| claude-haiku-5-5 | Low cost | 100% | 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: False-merge and false-split rates measured on a labelled sample before go-live and periodically; merges counted alone say nothing about quality.
How it is assessed: Answers a written question on the match quality measurement correctly
Related knobits
Builds on
Needed for
- R-306 Recognises an unmeasured false-merge rate
- R-503 Recognises a merge count used as a quality metric
- R-504 Recognises a duplicate rate measured after merging
- R-532 Adds a false-merge measure taken before merging
- R-552 Explains MDM quality metrics
All 16 knobits of Define the matching rules.
Source: Practice the stage 3 reference and review checklist rest on