Explains analytic rubric scoring
One score per criterion from that criterion's evidence, on its scale, with a reason; caps and totals applied afterwards (Standards 2014, ch. 6).
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 | 100% | Mastered |
| gpt-6-luna | Low cost | 99% | Not shown |
| 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: One score per criterion from that criterion's evidence, on its scale, with a reason; caps and totals applied afterwards (Standards 2014, ch. 6).
How it is assessed: Answers a written question on the analytic rubric scoring correctly
Related knobits
Needed for
- T-301 Recognises credit given for what an answer does not say
- T-302 Recognises an equivalent solution not credited
- T-306 Recognises a score off the rubric's scale
- T-307 Recognises a cap applied by the grader
- T-308 Recognises a grading reason citing no evidence
- T-309 Recognises a named failure mode left unreported
- T-310 Recognises a computed result taken on trust
- T-331 Scores each criterion from its own evidence in the answer
- T-332 Writes a grading reason citing the answer
All 16 knobits of Grade answers against a rubric.
Source: Practice the stage 3 reference and review checklist rest on