Falsify OÜ
Organisation
The Pre-Registered ML Manifest: an open specification for committing the success criteria of an AI evaluation (metric, threshold, dataset, seed and five further fields) to canonical bytes and a SHA...
The reference command line tool and library for PRML: canonicaliser, validator, verifier with dataset content checks, in-toto attestation output, and the lock/run/verdict/guard workflow that binds ...
JavaScript implementation of PRML, byte-equivalent to the Python, Go and Rust implementations. Byte equivalence across all four is enforced in CI against the shared conformance vectors on every pus...
Opt-in adapters that let existing evaluation tooling emit and verify PRML manifests without changing how teams work: an adapter for the UK AI Safety Institute's Inspect framework (listed in its off...
Organisation
Falsify OÜ
