Recognises a criterion no grader can decide
A criterion judged on impression, not on what the answer contains.
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 | 94% | Mastered |
| gemini-3.5-flash-lite | Low cost | 54% | Not shown |
| gpt-6-luna | Low cost | 94% | Mastered |
| claude-haiku-5-5 | Low cost | 94% | 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: What the answer must contain
How it is assessed: Flags the deficient case and passes the acceptable one. Deficient: I1 "Shows a good understanding of matching" Acceptable: "Names CU-03 and CU-08 as one person: same date of birth and address, name transliterated"
Also known as: Criterion no grader can decide
Related knobits
Builds on
All 15 knobits of Write and review test items.
Source: Review checklist 02-test-items, row 2 (Standards 4, 6)