15 citations · 16 across the 8 of their papers we have counts for
9 papers
What Does It Mean to Break a Distillation Defense?
Lena Libon, Pura Peetathawatchai, Michael Aerni +2
Black-box LLMs (accessible only via API) are vulnerable to distillation attacks, in which an attacker queries the model and trains a student on its outputs. A recent line of work p…
Large-scale online deanonymization with LLMs
Simon Lermen, Daniel Paleka, Joshua Swanson +3
We show that large language models can be used to perform at-scale deanonymization. With full Internet access, our agent can re-identify Hacker News users and Anthropic Interviewer…
Modal Aphasia: Can Unified Multimodal Models Describe Images From Memory?
Michael Aerni, Joshua Swanson, Kristina Nikolić +1
We present modal aphasia, a systematic dissociation in which current unified multimodal models accurately memorize concepts visually but fail to articulate them in writing, despite…
Apertus: Democratizing Open and Compliant LLMs for Global Language Environments
Project Apertus, Alejandro Hernández-Cano, Alexander Hägele +100
We present Apertus, a fully open suite of large language models (LLMs) designed to address two systemic shortcomings in today's open model ecosystem: data compliance and multilingu…
Membership Inference Attacks on Sequence Models
Lorenzo Rossi, Michael Aerni, Jie Zhang +1
Sequence models, such as Large Language Models (LLMs) and autoregressive image generators, have a tendency to memorize and inadvertently leak sensitive information. While this tend…
Measuring Non-Adversarial Reproduction of Training Data in Large Language Models
Michael Aerni, Javier Rando, Edoardo Debenedetti +3
Large language models memorize parts of their training data. Memorizing short snippets and facts is required to answer questions about the world and to be fluent in any language. B…
Evaluations of Machine Learning Privacy Defenses are Misleading
Michael Aerni, Jie Zhang, Florian Tramèr
Empirical defenses for machine learning privacy forgo the provable guarantees of differential privacy in the hope of achieving higher utility while resisting realistic adversaries.…
Strong inductive biases provably prevent harmless interpolation
Michael Aerni, Marco Milanta, Konstantin Donhauser +1
Classical wisdom suggests that estimators should avoid fitting noise to achieve good generalization. In contrast, modern overparameterized models can yield small test error despite…
Interpolation can hurt robust generalization even when there is no noise
Konstantin Donhauser, Alexandru Ţifrea, Michael Aerni +2
Numerous recent works show that overparameterization implicitly reduces variance for min-norm interpolators and max-margin classifiers. These findings suggest that ridge regulariza…