Writes a rubric criterion observing one knobit set
A criterion, or parts with points, each naming what the answer must contain and the item it grades
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 | 98% | Mastered |
| gemini-3.5-flash-lite | Low cost | 75% | Not shown |
| gpt-6-luna | Low cost | 98% | Mastered |
| claude-haiku-5-5 | Low cost | 95% | 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: A criterion, or parts with points, each naming what the answer must contain and the item it grades
How it is assessed: Each criterion or part maps to one knobit set and is decidable from the answer
Related knobits
Builds on
All 15 knobits of Write and review test items.
Source: Graded on its own by sa-item-authoring