README.md

---
title: README
emoji: 👁
colorFrom: gray
colorTo: indigo
sdk: static
pinned: false
---

<div class="grid lg:grid-cols-3 gap-x-4 gap-y-7">
	<div class="lg:col-span-3">
		<p class="text-xl font-semibold mb-2">The AI engine for business.</p>
		<p>
			<strong>Market Standard</strong> is a full-service technology studio. We design, build, and
			ship production AI, computer-vision, and LLM systems &mdash; from custom models and data
			pipelines to the <a href="https://www.marketstandard.app/" class="underline">Standard Suite</a>
			of developer tools. On Hugging Face we publish our open-source research, beginning with Panoptes.
		</p>
	</div>
	<div class="lg:col-span-3 flex items-start gap-x-4">
		<img src="https://huggingface.co/spaces/MarketStandard/README/resolve/main/images/panoptes.svg" alt="Panoptes logo" class="w-14 h-14 shrink-0" />
		<div>
			<p class="font-semibold text-lg mb-1">Panoptes &mdash; calibrated evidence for AI-text attribution</p>
			<p class="text-sm text-gray-800">
				Our flagship open-source release reframes AI-text detection as <em>evidence, not verdicts</em>.
				Rather than a binary &ldquo;AI or not&rdquo; label, Panoptes estimates <em>how much</em> of a
				document is AI-authored, tells you how much to trust that estimate, and measures whether the
				evidence still holds up when the population shifts.
			</p>
		</div>
	</div>
	<div class="lg:col-span-3">
		<p class="font-semibold mb-2">What it gives the community</p>
		<ul class="list-disc pl-5 space-y-1 text-sm text-gray-800">
			<li>A <strong>participation score</strong> (0&ndash;100% AI-authored) with split-conformal coverage &mdash; an estimate with valid uncertainty, not a false-precision verdict.</li>
			<li>A <strong>frozen, preregistered open-weights detector</strong> (three-seed DeBERTa-v3 ensemble) with verified hashes &mdash; a strong, reproducible baseline anyone can run.</li>
			<li>An <strong>evidence-transportability harness</strong> &mdash; cross-dataset leave-one-out and calibration-transfer matrices that quantify how scores and calibration generalize across MAGE, RAID, DeFactify, CoAuthor, M4GT, and EvoBench.</li>
			<li>A <strong>license-audited dataset registry and split manifest</strong>, so others can add cohorts and extend the corpus on solid legal footing.</li>
			<li>A <strong>preregistered protocol with signed cards</strong>, so every reported number is reproducible.</li>
		</ul>
	</div>
	<a href="https://www.marketstandard.app/" class="block overflow-hidden group">
		<div class="w-full h-36 rounded-lg flex flex-col items-center justify-center bg-gradient-to-b from-indigo-600 to-indigo-800 text-white px-4 text-center">
			<div class="text-2xl mb-1">&#127760;</div>
			<div class="font-semibold">marketstandard.app</div>
			<div class="text-xs text-indigo-100 mt-1">products &amp; platforms</div>
		</div>
		<div class="underline">Main site</div>
	</a>
	<a href="https://huggingface.co/MarketStandard/panoptes-neural-v1" class="block overflow-hidden group">
		<div class="w-full h-36 rounded-lg flex flex-col items-center justify-center bg-gradient-to-b from-slate-900 to-indigo-950 text-white px-4 text-center">
			<div class="text-2xl mb-1">&#128065;</div>
			<div class="font-semibold">panoptes-neural-v1</div>
			<div class="text-xs text-slate-300 mt-1">open-weights neural detector</div>
		</div>
		<div class="underline">Model on the Hub</div>
	</a>
	<a href="https://github.com/marketstandard/Panoptes" class="block overflow-hidden group">
		<div class="w-full h-36 rounded-lg flex flex-col items-center justify-center bg-gray-900 text-white px-4 text-center">
			<div class="text-2xl mb-1">&#128230;</div>
			<div class="font-semibold">marketstandard/Panoptes</div>
			<div class="text-xs text-gray-300 mt-1">source, protocol &amp; benchmark</div>
		</div>
		<div class="underline">GitHub &amp; paper</div>
	</a>
</div>