Gauthier Gidel

sadhana.anand@mila.quebec

Biographie

Je suis professeur adjoint au Département d’informatique et de recherche opérationnelle (DIRO) de l'Université de Montréal et membre académique principal de Mila – Institut québécois d’intelligence artificielle. J'ai obtenu une bourse Borealis AI destinée aux étudiant·e·s des cycles supérieurs et je suis actuellement titulaire d'une chaire en IA Canada-CIFAR. J'ai travaillé chez DeepMind et Element AI, et j'ai récemment été un visiteur de longue durée au Simons Institute de l’Université de Californie à Berkeley. Mes intérêts de recherche se situent à l'intersection de la théorie des jeux, de l'optimisation et de l'apprentissage automatique.

Étudiants actuels

Sadhana Anand

Maîtrise recherche - UdeM

yuan-chen.chang@mila.quebec

Yuan-Chen Chang

Doctorat - UdeM

David Dobre

Doctorat - UdeM

david-a.dobre@mila.quebec

damien.ferbach@mila.quebec

Nouha Dziri

Visiteur de recherche indépendant - N/A

nouha.dziri@gmail.com

Doctorat - UdeM

Co-superviseur⋅e :

Courtney Paquette

niklas.herbster@gmail.com

Niklas Herbster

Stagiaire de recherche - UdeM

Superviseur⋅e principal⋅e :

Irina Rish

marco.jiralerspong@mila.quebec

Marco Jiralerspong

Doctorat - UdeM

Zichu Liu

Doctorat - UdeM

Co-superviseur⋅e :

Ioannis Mitliagkas

zichu.liu@mila.quebec

Orel Mazor

Collaborateur·rice de recherche - ENS Lyon

orel.mazor@mila.quebec

andjela.mladenovic@mila.quebec

Andjela Mladenovic

Doctorat - UdeM

Co-superviseur⋅e :

Aaron Courville

mehrnaz.mofakhami@mila.quebec

Mehrnaz Mofakhami

Collaborateur·rice de recherche - UdeM

Co-superviseur⋅e :

Ioannis Mitliagkas

Collaborateur·rice de recherche - UdeM

ailin.rasteh@mila.quebec

Leo Schwinn

Visiteur de recherche indépendant - Technical Univeristy of Munich

leo.schwinn@mila.quebec

Tom Stanic

Stagiaire de recherche - UdeM

tom.stanic@mila.quebec

tommaso.tosato@mila.quebec

Tommaso Tosato

Postdoctorat - UdeM

Doctorat - UdeM

danilo.vucetic@mila.quebec

bora.yongacoglu@mila.quebec

Sophie Xhonneux

Doctorat - UdeM

Co-superviseur⋅e :

Collaborateur·rice alumni - N/A

Billets de blogue

13 juin 2024

Que nous apprennent les distributions des coefficients synaptiques au sujet de l’apprentissage dans le cerveau ?

par

Roman Pogodin

Jonathan Cornford

Arna Ghosh

Gauthier Gidel

Guillaume Lajoie

Blake Richards

Lire l'article

Publications

Activation Steering for Aligned Open-ended Generation without Sacrificing Coherence

Niklas Herbster

Martin Zborowski

Alberto Tosato

Tommaso Tosato

Alignment in LLMs is more brittle than commonly assumed: misalignment can be induced by adversarial prompts, benign fine-tuning, emergent mi… (voir plus)salignment, and goal misgeneralization. Recent evidence suggests that some misalignment behaviors are encoded as linear structure in activation space, making it tractable via activation steering, which could be used as a lightweight runtime defense. We implement three methods: Steer-With-Fixed-Coefficient (SwFC), which applies uniform additive steering, and two novel projection-aware methods, Steer-to-Target-Projection (StTP) and Steer-to-Mirror-Projection (StMP), that use a logistic regression decision boundary to selectively intervene only on tokens whose activations fall below the threshold. We evaluate these methods on two threat models, dishonesty and dismissiveness, using malicious system prompts as a controlled proxy for misalignment. We conduct our experiments on two architectures (Llama-3.3-70B-Instruct and Qwen3.6-27B). All methods substantially recover alignment. StTP and StMP preserve general capabilities (MMLU, MT-Bench, AlpacaEval) better than uniform steering. Finally, we show that our honesty steering generalizes to out-of-distribution scenarios: a single honesty direction extracted from the aligned model significantly raises scores on the MASK benchmark, suppresses deception in multi-agent settings (Among Us), doubles the hidden-behavior discovery rate on AuditBench, and restores honesty in an emergently misaligned model

2026-07-07

Conference on Language Modeling (accepté)

Accelerated and Stable Convergence with Anchored Optimistic Method

Jianxin You

We study first-order methods for solving monotone variational inequalities arising in min-max optimization. Classical approaches such as the… (voir plus) extragradient method rely on two gradient queries per iteration, which limits their analysis and applicability in the online and stochastic settings. We propose a family of Generalized Optimistic Methods with Anchoring (GOMA), which combine two-time-scale optimistic updates with an anchoring term inspired by Halpern iteration. In the deterministic setting, GOMA achieves the optimal accelerated last-iterate rate

2026-06-18

arXiv (prépublication)

arxiv.org

Why Open Source? A Game-Theoretic Analysis of the AI Race

Andjela Mladenovic

Aaron Courville

In recent years, with the advancement of frontier AI, we have observed certain dynamics in open-sourcing and closed-sourcing decisions. We p… (voir plus)ropose a game-theoretic model to analyze these dynamics in the current landscape of the AI race. Our model builds on an R&D race framework under a winner-takes-all setting, and it accounts for the cases where the players' actions can be either discrete or continuous (i.e., partial open-sourcing, such as open weights). We show that determining the existence of a discrete pure non-trivial Nash equilibrium is NP-hard in general but that we can transform the discrete Nash existence computation into a MIP (Mixed-Integer Programming) problem, making it tractable for small instances using a standard MIP solver. Next, we show the existence and tractability of pure Nash equilibria in the continuous version of our problem, leveraging standard convex analysis results, and constructing an equivalent MIP formulation. Throughout this work, we leverage both our main technical results as well as surrounding technical analysis, to derive socially relevant insights that we believe can serve both to understand already existing decisions and dynamics and to potentially inform new policies.

2026-04-16

arXiv (prépublication)

arxiv.org

Soft Mellowmax Monte Carlo Planning

Danilo Vucetic

Soft mellowmax (SMM) recently emerged as an alternative operator in Q-learning, achieving impressive performance in games and scientific dis… (voir plus)covery tasks. Despite SMM's ability to achieve high returns and its enticing robustness, diversity, and sample efficiency characteristics, SMM has not yet been translated into a Monte Carlo tree search algorithm. To address this gap, a soft mellowmax-based Monte Carlo tree search algorithm, SMM-TS, is proposed and theoretically justified. It is empirically demonstrated that SMM-TS converges significantly faster than other tree search methods in synthetic environments, while maintaining competitive performance in games. The fast convergence of SMM-TS makes recursive self-improvement loops more scalable, while the stability gained via planning and the robustness of the operator make SMM-TS more practical for agents operating in uncertain and changing environments.

2026-03-04

RSI @ International Conference on Learning Representations (poster)

Beyond Reward Maximization: Evaluating the Diversity of Trajectories in Reinforcement Learning with Temporal Vendi Score

In domains such as scientific discovery and automated design using reinforcement learning (RL), the final task of an agent should extend bey… (voir plus)ond maximising a single scalar reward; it requires identifying diverse sets of high-quality trajectories to uncover distinct solutions that can provide novel insights on how to solve the problems of interest and transfer robustly from simulation to the real world. However, the RL literature currently lacks a holistic, domain-agnostic standard for measuring trajectory diversity. Existing metrics have been developed to improve exploration at training time but not to evaluate and compare diversity induced by different agents, rendering cross-method comparisons inconsistent and challenging. To address this, we introduce the Temporal Vendi Score (TVS), a novel metric designed to evaluate the diversity of an RL agent by computing the entropy of the eigenvalues' similarity matrix of sampled trajectories. Unlike previous approaches, our metric captures the behavioural diversity of trajectories by accounting for both the sequential nature of state visitations and the temporal structure of the underlying MDP, rather than relying on order-agnostic state comparisons. We validate the TVS on simple environments where we can control the number of different ways a problem can be solved, demonstrating that it provides a more robust, semantically meaningful ranking of diversity than standard baselines. We then show that our metric can scale to a high-dimensional, continuous environment.

2026-03-01

LLA @ International Conference on Learning Representations (poster)

Logarithmic-time Schedules for Scaling Language Models with Momentum

In practice, the hyperparameters …

2026-02-04

arXiv (prépublication)

arxiv.org

Accelerated and Stable Convergence with Anchored Generalized Optimistic Method

Jianxin You

We study first-order methods for solving monotone variational inequalities arising in min-max optimization. Classical approaches such as the… (voir plus) extragradient method rely on two gradient queries per iteration, which limits their analysis and applicability in the online and stochastic settings. We propose a family of Generalized Optimistic Methods with Anchoring (GOMA), which combine two time-scale optimistic updates with an anchoring term inspired by Halpern iteration. In particular, we show that for monotone Lipschitz operators, GOMA achieves an accelerated last-iterate convergence rate of

2025-12-31

International Conference on Machine Learning (Accept (regular))

A Coin Flip for Safety: LLM Judges Fail to Reliably Measure Adversarial Robustness

Leo Schwinn

Moritz Ladenburger

Tim Beyer

Mehrnaz Mofakhami

Stephan Günnemann

Automated \enquote{LLM-as-a-Judge} frameworks have become the de facto standard for scalable evaluation across natural language processing. … (voir plus)For instance, in safety evaluation, these judges are relied upon to evaluate harmfulness in order to benchmark the robustness of safety against adversarial attacks. However, we show that existing validation protocols fail to account for substantial distribution shifts inherent to red-teaming: diverse victim models exhibit distinct generation styles, attacks distort output patterns, and semantic ambiguity varies significantly across jailbreak scenarios. Through a comprehensive audit using 6642 human-verified labels, we reveal that the unpredictable interaction of these shifts often causes judge performance to degrade to near random chance. This stands in stark contrast to the high human agreement reported in prior work. Crucially, we find that many attacks inflate their success rates by exploiting judge insufficiencies rather than eliciting genuinely harmful content. To enable more reliable evaluation, we propose ReliableBench, a benchmark of behaviors that remain more consistently judgeable, and JudgeStressTest, a dataset designed to expose judge failures. (Data in supplement).

2025-12-31

International Conference on Machine Learning (Accept (regular))

Position: LLM-Safety Evaluations Lack Robustness

Tim Beyer

Sophie Xhonneux

Simon Geisler

Leo Schwinn

Stephan Günnemann

In this position paper, we argue that current safety alignment research efforts for large language models are hindered by many intertwined s… (voir plus)ources of noise, such as small datasets, methodological inconsistencies, and unreliable evaluation setups. This can, at times, make it impossible to evaluate and compare attacks and defenses fairly, thereby slowing research progress. We systematically analyze the LLM safety evaluation pipeline, covering dataset curation, optimization strategies for automated red-teaming, response generation, and response evaluation using LLM judges. At each stage, we identify key issues and highlight their practical impact. We also propose a set of guidelines for reducing noise and bias in evaluations of future attack and defense papers. Lastly, we offer an opposing perspective, highlighting practical reasons for existing limitations. We believe that addressing the outlined problems in future research will improve the field’s ability to generate easily comparable results and make measurable progress.

2025-12-31

International Conference on Machine Learning (Accept (regular))

Dimension-adapted Momentum Outscales SGD

We investigate scaling laws for stochastic momentum algorithms with small batch on the power law random features model, parameterized by dat… (voir plus)a complexity, target complexity, and model size. When trained with a stochastic momentum algorithm, our analysis reveals four distinct loss curve shapes determined by varying data-target complexities. While traditional stochastic gradient descent with momentum (SGD-M) yields identical scaling law exponents to SGD, dimension-adapted Nesterov acceleration (DANA) improves these exponents by scaling momentum hyperparameters based on model size and data complexity. This outscaling phenomenon, which also improves compute-optimal scaling behavior, is achieved by DANA across a broad range of data and target complexities, while traditional methods fall short. Extensive experiments on high-dimensional synthetic quadratics validate our theoretical predictions and large-scale text experiments with LSTMs show DANA's improved loss exponents over SGD hold in a practical setting.

2025-12-02

Conference on Neural Information Processing Systems (Accept (spotlight))

Tight Lower Bounds and Improved Convergence in Performative Prediction

Performative prediction is a framework accounting for the shift in the data distribution induced by the prediction of a model deployed in th… (voir plus)e real world. Ensuring rapid convergence to a stable solution where the data distribution remains the same after the model deployment is crucial, especially in evolving environments. This paper extends the Repeated Risk Minimization (RRM) framework by utilizing historical datasets from previous retraining snapshots, yielding a class of algorithms that we call Affine Risk Minimizers and enabling convergence to a performatively stable point for a broader class of problems. We introduce a new upper bound for methods that use only the final iteration of the dataset and prove for the first time the tightness of both this new bound and the previous existing bounds within the same regime. We also prove that utilizing historical datasets can surpass the lower bound for last iterate RRM, and empirically observe faster convergence to the stable point on various performative prediction benchmarks. We offer at the same time the first lower bound analysis for RRM within the class of Affine Risk Minimizers, quantifying the potential improvements in convergence speed that could be achieved with other variants in our framework.

2025-12-02

Advances in Neural Information Processing Systems (Accept (poster))

Self-Play Q-Learners Can Provably Collude in the Iterated Prisoner's Dilemma

Quentin Bertrand

Juan Agustin Duque

Emilio Calvano

A growing body of computational studies shows that simple machine learning agents converge to cooperative behaviors in social dilemmas, such… (voir plus) as collusive price-setting in oligopoly markets, raising questions about what drives this outcome. In this work, we provide theoretical foundations for this phenomenon in the context of self-play multi-agent Q-learners in the iterated prisoner’s dilemma. We characterize broad conditions under which such agents provably learn the cooperative Pavlov (win-stay, lose-shift) policy rather than the Pareto-dominated “always defect” policy. We validate our theoretical results through additional experiments, demonstrating their robustness across a broader class of deep learning algorithms.

2025-10-05

Proceedings of the 42nd International Conference on Machine Learning (publié)