Recognises an automatic check contradicting its criterion
A checks.py check stricter or looser than the criterion it is linked to.
Can AI do this?
Not settled yet The benchmark tested this, but the models' answers do not settle yet whether they master it. 0 of 4 models in the current benchmark master it.
| Model | Tier | Probability of mastery | Status |
|---|---|---|---|
| gemini-4-argon | Flagship | 83% | Not shown |
| gemini-3.5-flash-lite | Low cost | 37% | Not shown |
| gpt-6-luna | Low cost | 36% | Not shown |
| claude-haiku-5-5 | Low cost | 83% | 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: The check aligned or made advisory
How it is assessed: Flags the deficient case and passes the acceptable one. Deficient: key_thresholds_absolute fails a percentage the criterion accepts Acceptable: The check tests what the criterion says, or is advisory
Also known as: Check contradicting its criterion
Related knobits
Builds on
All 15 knobits of Write and review test items.
Source: Review checklist 02-test-items, row 9 (Practice)