Explains targets and their rationale
A target has a rationale traceable to the goal, a supervisor, a benchmark or a baseline, and never contradicts the goal it measures.
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 | 99% | Mastered |
| gemini-3.5-flash-lite | Low cost | 59% | Not shown |
| gpt-6-luna | Low cost | 99% | Mastered |
| claude-haiku-5-5 | Low cost | 99% | 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 target has a rationale traceable to the goal, a supervisor, a benchmark or a baseline, and never contradicts the goal it measures.
How it is assessed: Answers a written question on the metric target correctly
Related knobits
Builds on
Needed for
- G-602 Recognises a governance metric with no target
- G-605 Recognises a metric target with no rationale
- G-612 Recognises a metric target inconsistent with its goal
All 21 knobits of Build the governance scorecard.
Source: Practice the stage 6 reference and review checklist rest on