Congratulations to Augustin Godinot, Erwan Le Merrer, and Francois Taiani for their paper entitled “Queries, Representation & Detection: the next 100 model fingerprinting schemes” accepted at AAAI 2025, a flagship conference in artificial intelligence.
Comparing learned models is a fundamental task in Machine Learning (ML). In this paper, they studied model fingerprinting techniques that compare the output of two ML models, A and B, to answer the question: has B stolen knowledge from A? The evaluation shows that a simple baseline that we introduce performs on par with existing state-of-the-art fingerprints, which, on the other hand, are much more complex. This is joint work with Camilla Penzo of the PEReN team and Gilles Tredan, Research Director at the LAAS CNRS laboratory.
