Sentient Foundation launches open source Fingerprinting Leaderboard
The leaderboard scores verification reliability and accuracy cost alongside removal resistance, which most model provenance work ignores even though a mark you cannot read back is worthless. Teams building open weight models in ASEAN can use it to pick a fingerprinting method before release. It compares directly against conventional watermarking and provenance approaches like C2PA metadata.
Sentient Foundation released a free, open source benchmark called the Fingerprinting Leaderboard that ranks methods for proving an AI model was copied. Researchers submit fingerprinting methods, which Sentient embeds into its own test models before attempting removal and ranking results on a public website. The underlying test code, mlprints, is also open sourced. Fingerprinting trains a model to return unlikely answers to secret prompts so owners can verify a suspected copy through its API alone. Sentient's own Perinucleus method embedded 24,576 fingerprints in Llama-3.1-8B without degrading utility, and the marks survived standard fine-tuning.
Good news for anyone trying to prove a model is theirs.
Sentient Foundation released the Fingerprinting Leaderboard, a free, open source benchmark that tests fingerprinting methods, meant to prove an AI model was copied from yours.
Researchers submit a method for hiding an ownership mark inside a model, and Sentient puts that mark into its own test models, tries to strip it out, then ranks the method on a public website.
The code that runs these tests, called mlprints, is released too.
A fingerprint trains a model to return unlikely answers to secret prompts, so an owner can test a suspected copy through its API alone.
But a fingerprint that survives every attack is still useless if the owner can’t read it back reliably.
Sentient Foundation’s Fingerprinting Leaderboard scores that verification step directly, alongside how well the mark resists removal and how much model skill it costs.
Sentient’s Perinucleus method embedded 24,576 fingerprints in Llama-3.1-8B without degrading utility, is also competing on the leaderboard, and the fingerprints survived standard fine-tuning.
