Skip to yearly menu bar Skip to main content


Invited Talk
in
Workshop: The Impact of Memorization on Trustworthy Foundation Models

Invited Talk 1: Reza Shokri - On Testing Memorization in AI: From Brute-Force Methods to Robust Statistical Tests

[ ]
Sat 19 Jul 9:30 a.m. PDT — 10 a.m. PDT

Abstract:

In this talk, the speaker will explore a range of techniques for detecting and analyzing memorization in machine learning models. The presentation will cover proof-of-concept brute-force extraction attacks, high-power and range-based membership inference attacks, as well as recent advances in dataset inference methods. The strengths and limitations of each approach will be discussed, offering a comprehensive perspective on the current landscape of memorization testing.

Chat is not available.