Paste a public URL or upload a file. We compute the LabelSets Quality Score across 7–14 dimensions, sign it with our production Ed25519 key, and return a procurement-grade certificate you can attach to vendor reviews. No selling, no listing — just the score.
Every cert is a self-contained, cryptographically-signed JSON envelope. Auditors and AI risk teams verify it offline against our published public key — no LabelSets account or login required on their end.
0–100 LQS composite plus the underlying dimensions: schema integrity, label production, modality fit, license clarity, contamination risk, and more.
SHA-256 cert hash + base64 signature. Re-verify against our published public key with one crypto.verify() call. No request to our server required.
Bronze → Silver → Gold → Platinum tiers with explicit confidence scores. URL-derived certs flag as proxy (cap 85); file-based certs run the full 14-dim engine for full procurement weight.
No metadata to fill out, no listing to maintain, no fee. The same scoring engine that gates our $549–$799 marketplace listings runs against whatever dataset you point at it.
Paste a HuggingFace or Zenodo URL, or upload a file (signed in).
Our scorer derives signals across 7–14 dimensions: schema, modality, label production, license, contamination, downstream projection.
The payload is canonicalized (sorted-key JSON), hashed (SHA-256), and signed with our production Ed25519 key.
Auditors run crypto.verify() against our public key to confirm the cert hasn't been tampered with. No round-trip to LabelSets needed.
If your dataset clears Gold or Platinum, you can flip it into a paid listing in under two minutes. The cert you just got carries straight over — same hash, same signature, same audit trail.
Every dataset on LabelSets ships with a signed LQS cert — Gold or Platinum tier, contamination-checked against 40+ public evals, with 90% conformal prediction intervals on downstream F1. Plug it into your model package without the rate-and-clean cycle.