Recognises paraphrased data requirements
Two requirements that say the same thing in different words, often from different stakeholders or documents. The script finds close wording; a person finds the same need described differently.
Can AI do this?
People still lead Few or none of the tested AI models showed they master this. 0 of 10 models master it.
| Model | Tier | Probability of mastery | Status |
|---|---|---|---|
| gemini-4-argon | Flagship | 71% | Not shown |
| gemini-3.1-pro-preview | Flagship | 38% | Not shown |
| gpt-6-astra | Flagship | 39% | Not shown |
| claude-fable-5-1 | Flagship | 38% | Not shown |
| gemini-3.8-flash | Mid tier | 36% | Not shown |
| gpt-6.1-sol | Mid tier | 36% | Not shown |
| claude-sonnet-5-5 | Mid tier | 37% | Not shown |
| gemini-3.5-flash-lite | Low cost | 38% | Not shown |
| gpt-6-luna | Low cost | 36% | Not shown |
| claude-haiku-5-5 | Low cost | 37% | 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 paraphrase pair and the one to retire
How it is assessed: Flags the deficient case and passes the acceptable one. Deficient: DR-011 "…the date on which the payment was instructed by the payer" alongside DR-006; the first half of DR-001 says the same again Acceptable: One requirement; the second stakeholder or document recorded in rationale or a YAML comment, and the others retired with superseded_by.
Also known as: Paraphrase (data requirement)
Related knobits
Builds on
Usually followed by
All 47 knobits of Capture data requirements.
Source: Review checklist 01-requirements, row 22; 29148 5.2.6 (consistent); REQ-011