GapAnalyzer is built on published, checkable research into how well language models and human experts actually assess maritime compliance documents. This is where that work lives.
Two DP experts and an independent maritime consultancy assessed the same operations manual against the same 99-item checklist, and agreed on roughly half the items. What expert disagreement means for every accuracy claim in compliance AI, with the full interactive data.
The data on its own: pick your reference and watch eight model rankings invert, see the leniency spectrum, and inspect all eleven assessors' verdicts on every checklist item.