LQS v3.1 — an open methodology for rating AI training data (multi-oracle consensus + signed certificates) [P]

In three linesLQS v3.1 is an open-source methodology for rating AI training data quality. It uses 19 dimensions (label correctness, contamination, equity, etc.), multi-oracle consensus (7 oracles) with real-world outcome recalibration, and offline-verifiable Ed25519 certificates. Free public index with 263 scored datasets.

Summary generated by Claude — human-verified