Recognises a threshold at odds with its rule
100 % on data known to fail, which gets the check switched off within a week; or below 100 % on a rule that is absolute — a key, a reference, an alethic rule — which turns a corrupt key into a dashboard. A threshold below 100 % with no record of who agreed it is a rule nobody agreed to.
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 | 83% | Not shown |
| gemini-3.1-pro-preview* | Flagship | 100% | Mastered |
| gpt-6-astra* | Flagship | 100% | Mastered |
| claude-fable-5-1* | Flagship | 100% | Mastered |
| gemini-3.8-flash* | Mid tier | 100% | Mastered |
| gpt-6.1-sol* | Mid tier | 100% | Mastered |
| claude-sonnet-5-5* | Mid tier | 99% | Mastered |
| gemini-3.5-flash-lite | Low cost | 35% | Not shown |
| gpt-6-luna | Low cost | 84% | Not shown |
| claude-haiku-5-5 | Low cost | 82% | Not shown |
* From the earlier benchmark of 2026-10: 10 data-modelling test tasks, 3 runs per task, graded by claude-opus-5-5. These models have not taken the current 123 tasks yet; the two methods differ, so compare their figures with care. They are left out of the “most models” counts and the best-model picks.
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: The threshold and the value the rule implies
How it is assessed: Flags the deficient case and passes the acceptable one. Deficient: uniqueness on account_identifier at threshold: "98%"; or a new completeness rule at 100 % over ten years of migrated data that is 92 % complete Acceptable: Keys and references at 100 %; legacy-tolerant rules at the measured rate with an agreed date to tighten, and the agreement in comment (DQR-021)
Also known as: Threshold at odds with the rule (data quality rule)
Related knobits
Builds on
All 30 knobits of Derive data quality rules.
Source: Review checklist 03-dq-rules (data-quality), row 20; Loshin 2011; DMBOK2; formerly data-modelling K-720