15 citations · 47 across the 20 of their papers we have counts for
9 papers · 1 filter
High-probability zeroth-order online convex optimisation beyond Euclidean geometry
David Janz, El-Mahdi El-Mhamdi, Arya Akhavan
We study online convex optimisation with -Lipschitz losses, -regularised FTRL, and randomised two-point finite-difference gradient estimators based on cone-measure…
Winter Soldier: Backdooring Language Models at Pre-Training with Indirect Data Poisoning
Wassim Bouaziz, Mathurin Videau, Nicolas Usunier +1
The pre-training of large language models (LLMs) relies on massive text datasets sourced from diverse and difficult-to-curate origins. Although membership inference attacks and hid…
On Monotonicity in AI Alignment
Gilles Bareilles, Julien Fageot, Lê-Nguyên Hoang +4
Comparison-based preference learning has become central to the alignment of AI models with human preferences. However, these methods may behave counterintuitively. After empiricall…
The Strong, Weak and Benign Goodhart's law. An independence-free and paradigm-agnostic formalisation
Adrien Majka, El-Mahdi El-Mhamdi
Goodhart's law is a famous adage in policy-making that states that ``When a measure becomes a target, it ceases to be a good measure''. As machine learning models and the optimisat…
Targeted Data Poisoning for Black-Box Audio Datasets Ownership Verification
Wassim Bouaziz, El-Mahdi El-Mhamdi, Nicolas Usunier
Protecting the use of audio datasets is a major concern for data owners, particularly with the recent rise of audio deep learning models. While watermarks can be used to protect th…
On the Byzantine Fault Tolerance of signSGD with Majority Vote
Emanuele Mengoli, Luzius Moll, Virgilio Strozzi +1
In distributed learning, sign-based compression algorithms such as signSGD with majority vote provide a lightweight alternative to SGD with an additional advantage: fault tolerance…
A Case for Specialisation in Non-Human Entities
El-Mahdi El-Mhamdi, Lê-Nguyên Hoang, Mariame Tighanimine
With the rise of large multi-modal AI models, fuelled by recent interest in large language models (LLMs), the notion of artificial general intelligence (AGI) went from being restri…
Approaching the Harm of Gradient Attacks While Only Flipping Labels
Abdessamad El-Kabid, El-Mahdi El-Mhamdi
Machine learning systems deployed in distributed or federated environments are highly susceptible to adversarial manipulations, particularly availability attacks -- rendering the t…
Stop treating `AGI' as the north-star goal of AI research
Borhane Blili-Hamelin, Christopher Graziul, Leif Hancox-Li +13
The AI research community plays a vital role in shaping the scientific, engineering, and societal goals of AI research. In this position paper, we argue that focusing on the highly…