Yoshua Bengio

jamal.abouhaibeh@mila.quebec

Biography

*For media requests, please write to medias@mila.quebec.

For more information please contact Cassidy MacNeil, Senior Assistant and Operation Lead at cassidy.macneil@mila.quebec.

Yoshua Bengio is recognized worldwide as a leading expert in AI. He is most known for his pioneering work in deep learning, which earned him the 2018 A.M. Turing Award, “the Nobel Prize of computing,” with Geoffrey Hinton and Yann LeCun.

Bengio is a full professor at Université de Montréal, and the founder and scientific advisor of Mila – Quebec Artificial Intelligence Institute. He is also a senior fellow at CIFAR and co-directs its Learning in Machines & Brains program, serves as special advisor and founding scientific director of IVADO, and holds a Canada CIFAR AI Chair.

In 2019, Bengio was awarded the prestigious Killam Prize and in 2022, he was the most cited computer scientist in the world by h-index. He is a Fellow of the Royal Society of London, Fellow of the Royal Society of Canada, Knight of the Legion of Honor of France and Officer of the Order of Canada. In 2023, he was appointed to the UN’s Scientific Advisory Board for Independent Advice on Breakthroughs in Science and Technology.

Concerned about the social impact of AI, Bengio helped draft the Montréal Declaration for the Responsible Development of Artificial Intelligence and continues to raise awareness about the importance of mitigating the potentially catastrophic risks associated with future AI systems.

Current Students

Jamal Abou Haibeh

Collaborating Alumni - McGill University

Collaborating researcher - Cambridge University

Principal supervisor :

David Rolnick

berkes.anais@mila.quebec

sanghyeok.choi@mila.quebec

Rim Assouel

PhD - Université de Montréal

assouelr@mila.quebec

Xiaoyin Chen

PhD - Université de Montréal

xiaoyin.chen@mila.quebec

Sanghyeok Choi

Collaborating researcher - KAIST

PhD - Université de Montréal

aniket.didolkar@mila.quebec

PhD - Université de Montréal

Co-supervisor :

Guillaume Lajoie

eric.elmoznino@mila.quebec

PhD - Université de Montréal

lena-nehale.ezzine@mila.quebec

Jean-Pierre Falet

PhD - Université de Montréal

jean-pierre.falet@mila.quebec

PhD

pietro.greiner@mila.quebec

Mohsin Hasan

PhD - Université de Montréal

mohsin.hasan@mila.quebec

thomas.jiralerspong@mila.quebec

Moksh Jain

PhD - Université de Montréal

moksh.jain@mila.quebec

PhD - Université de Montréal

Principal supervisor :

Guillaume Lajoie

Collaborating Alumni - Université de Montréal

younesse.kaddar@mila.quebec

Minsu Kim

Collaborating researcher - Université de Montréal

minsu.kim@mila.quebec

Hyeonah Kim

Postdoctorate - Université de Montréal

Principal supervisor :

Alex Hernández-García

hyeonah.kim@mila.quebec

Jeongjae Lee

Research Intern - KOREA ADVANCED INSTITUTE OF SCIENCE AND TECHNOLOGY

jeongjae.lee@mila.quebec

Collaborating researcher

Song LIU

Collaborating researcher - s.o.

liusong299@gmail.com

cristian-dragos.manta@mila.quebec

Cristian Dragos Manta

PhD - Université de Montréal

Co-supervisor :

Dhanya Sridhar

Sarthak Mittal

PhD - Université de Montréal

Principal supervisor :

Independent visiting researcher - Université de Montréal

nguyenph@mila.quebec

Ali Parviz

Collaborating researcher - Ying Wu Coll of Computing

ali.parviz@mila.quebec

Lena Podina

Collaborating researcher - University of Waterloo

Principal supervisor :

David Rolnick

elena.podina@mila.quebec

jarrid.rector-brooks@mila.quebec

Jarrid Rector-Brooks

PhD - Université de Montréal

Danyal REHMAN

Postdoctorate - Université de Montréal

danyal.rehman@mila.quebec

oliver.richardson@mila.quebec

Oli RICHARDSON

Postdoctorate - Université de Montréal

Camille Rochefort-Boulanger

PhD - Université de Montréal

Principal supervisor :

Postdoctorate - Université de Montréal

Co-supervisor :

Alex Hernández-García

divya.sharma@mila.quebec

Mélisande Astrid Crystal Teng

Collaborating Alumni - Université de Montréal

Co-supervisor :

Alex Tong

Collaborating Alumni - Université de Montréal

alexander.tong@mila.quebec

Collaborating researcher - University of Cambridge

oszkar.urban@mila.quebec

donna.vakalis@mila.quebec

Donna Vakalis

Collaborating Alumni - Université de Montréal

Co-supervisor :

David Rolnick

PhD - Université de Montréal

Principal supervisor :

Glen Berseth

siddarth.venkatraman@mila.quebec

Collaborating researcher

joseph.viviano2@mila.quebec

Collaborating researcher - Université de Montréal

esmeralda.whitammer@mila.quebec

Collaborating Alumni

omar.younis@mila.quebec

Collaborating researcher - Université de Montréal

taeyoung.yun@mila.quebec

Skipper: Combining Spatial and Temporal Abstraction for Better Generalization

Tianyu Zhang

PhD - Université de Montréal

tianyu.zhang@mila.quebec

PhD - McGill University

Principal supervisor :

Harry Zhao

Collaborating Alumni - McGill University

Principal supervisor :

Blog Posts

Generic thumbnail for Mila Blog articles.

February 22, 2024

Mingde Harry Zhao

Safa Alver

Harm van Seijen

Romain Laroche

Doina Precup

Yoshua Bengio

Scaling in the Service of Reasoning & Model-Based ML

April 4, 2023

Yoshua Bengio

Edward J. Hu

A collaboration between Mila and Relation Therapeutics to discover novel synergistic combinations of drugs in vitro

March 23, 2022

Paul Bertin

Jake P. Taylor-King

Yoshua Bengio

March 15, 2022

Generative Flow Networks

Yoshua Bengio

Publications

Adaptive Order Policies for Masked Diffusion

Jama Hussein Mohamud

Mohsin Hasan

Mirco Ravanelli

Masked diffusion models have seen great success in capturing data distributions over discrete sequences in domains such as text and proteins… (see more). These models generate data by iteratively unmasking tokens starting from a fully masked sequence, with the unmasking order typically chosen at random or using a heuristic based on denoiser probabilities. In this work, we propose a scheme for learning the unmasking order using an additional lightweight policy network on top of a diffusion model. Our proposed loss reweights terms in the masked diffusion loss according to policy probabilities, and results in a policy that prefers positions where the denoiser is more likely to be correct. We study this loss in two settings: (i) training solely the policy while using a frozen pre-trained denoiser, and (ii) training the policy and denoiser jointly with the weighted loss to allow for mutual adaptation. We demonstrate that our approach outperforms common heuristics on problems that are sensitive to token ordering, such as combinatorial problems, proteins as well as various coding and language tasks.

2026-12-30

OpenReview (published)

openreview.net

Extended generalization and improved safety with the System‐2 features of the Scientist AI

Abstract Neural networks usually answer queries with a single forward pass—an analog of Kahneman's System 1. In unfamiliar or high‐stake… (see more)s regimes, this can be brittle and overconfident. The Scientist AI framework augments a fast Predictor trained to approximate the Bayesian posterior with System 2 inference: structured probabilistic reasoning over intermediate natural‐language statements and causal hypotheses. The units that are generated and related via probabilistic dependencies are not the words or tokens but these Boolean statements about arbitrary properties of the world. Because inference‐time compute is spent only on the query at hand, marginalizing over latent statements can reduce computational uncertainty, produce better calibrated conditional probabilities and confidence intervals, and generate explanations as explicit argument graphs. With data transformed into these statements processed to make a syntactic difference between communication acts and factual hypotheses and the Predictor trained to be consequence‐invariant —insensitive to downstream effects of its outputs (Fornasiere et al., 2026)—this epistemic caution supports a safety case: the system is not imitating humans, has less incentive to acquire implicit goals, and is better suited as a guardrail for untrusted agents.

2026-10-21

AI Magazine (published)

Evaluating Predictor-Corrector Dependence in Hard-Constrained Climate Emulators

Tianyu Zhang

Xinyu Wang

Zhenghan Tai

Xiao-Wen Chang

Climate emulators often enforce physical constraints by correcting each state predicted by a neural network, but evaluating only the correct… (see more)ed forecast conflates predictor skill with the effect of the correction. We introduce an evaluation framework that separates these contributions by comparing corrected forecasts, pre-correction predictions along corrected trajectories, and autonomous rollouts with correction disabled. We further cross correction during training and inference to distinguish its immediate effect from how it shapes the learned predictor, and use matched nonphysical controls to test whether improvements are specific to the physical correction. In ACE2, correction substantially improves physical consistency. In a three-seed ACE2 pilot, models trained with correction produce less accurate uncorrected precipitation forecasts at short horizons than models trained without correction, but more accurate forecasts at 10 and 30 days. This contradicts our preregistered hypothesis that the deficit would grow with forecast horizon. In a separate nine-seed experiment, directly supervising the uncorrected output reverses its precipitation forecast deficit at 72 hours while maintaining the accuracy of corrected forecasts at 24 hours. In ConCerNet's heat-equation system, projection-trained models have higher uncorrected trajectory-based vector-field RMSE than models trained without projection, while restoring the physical projection recovers accuracy far more effectively than a matched nonphysical control. Together, these results show that hard corrections can reshape the predictors trained alongside them, creating dependencies that are not visible from corrected forecasts alone. Hard-constrained emulators should therefore be evaluated by separately measuring predictor accuracy, correction effects, and their interaction during training.

2026-10-15

arXiv (preprint)

GCAM-Gym: A Benchmark for Agent-Based Climate Policy Search

Tianyu Zhang

Zhenghan Tai

Xinyu Wang

Irina Rish

Comparing climate-policy search methods requires accounting for differences in policy control, available observations, feedback timing, and … (see more)simulator budgets. We introduce GCAM-Gym, a Gymnasium-compatible environment that supports full-policy evaluation and stepwise control through a common interface and exposes the components of the policy objective. We compare seven search and control systems in three settings: target attainment and cost, cost-temperature trade-offs, and one-step delayed feedback. Each system is evaluated with ten seeds per setting, for a total of 3,210 attempted GCAM evaluations. Four systems share a budget of 24 attempts per seed; among these, CMA-ES achieves the highest interquartile mean (IQM) of the best returns from completed policy evaluations in all three settings. We release the evaluation archive and analysis tools to reconstruct the reported comparisons without running GCAM.

2026-10-15

arXiv (preprint)

Defining and Generating Compositional Structure

Eric Elmoznino

Tejas Kasetty

Blaise Agüera y Arcas

Guillaume Lajoie

Compositional structure is pervasive in nature and across scales: atoms combine into molecules, molecules into cells, and cells into organis… (see more)ms, while further compositional hierarchies appear within each level. More broadly, compositionality, the reuse and recombination of parts into novel structures, is a fundamental source of interesting complexity in both natural and artificial systems. Yet despite its importance, and despite our apparent ease in recognizing it, mathematical accounts capturing what we intuitively mean by compositional structure remain limited. Here, we fill this gap by formalizing a simple idea: compositional structure emerges when parts, considered together, reveal shared structure that is not captured when those parts are modeled separately. We express this idea using algorithmic information theory, the mathematics of information and compression, which provides a natural way to identify structure in arbitrary data without specifying in advance what that structure must look like. Applied across scales, our definition informs where and how compositional structure emerges within an object. Using tractable approximations, we then invert the framework to generate compositional objects: we specify the desired form of their compositional structure without prescribing their particular content, and optimize directly over strings and images. The resulting objects exhibit readily recognizable hierarchical reuse and part-whole organization, suggesting that the definition captures compositional structure rather than merely complexity, regularity, or shared information. Finally, we discuss implications for machine learning, including a new perspective on depth in neural architectures and principles for constructing compositional data curricula capable of supporting open-ended learning.

2026-10-13

arXiv (preprint)

The Smart Buildings Challenge: Learning Foundation Models for Real-World Optimization at Scale

Judah Goldfeder

Ayush Rai

Philippe M. Wyder

Na Li

Victoria Dean

Donna Vakalis

Jiong Lin

Bing Dong

Michael Rossetti

Nathan Kutz

Sohei Arisaka

John Sipple

Hod Lipson

Buildings cause nearly 40\% of carbon emissions in the US. Of total building emissions, HVAC systems are responsible for approximately 40-60… (see more)\% and account for 15\% of global energy consumption. Despite advances in artificial intelligence, most buildings still operate as inefficiently as they did decades ago. Building optimization, perhaps more so than any other optimization problem, has enormous potential for Machine Learning (ML) driven impact on climate change, in both the short and long term. One of the biggest obstacles for building optimization is the lack of policy transfer. Most building control systems, whether simulation or learning based, don't generalize: they only work well for the building they were developed for, but not for other buildings. To solve this problem, we propose a challenge for the development of building foundation models that easily adapt to new buildings with minimal configuration and thereby enable building optimization to scale across the globe. Participants will build control policies on a set of simulated buildings, and have their policies evaluated on an unseen set of buildings. The buildings vary in size, layout, weather, material properties, and HVAC system. Successful outcomes from this initiative promise a massive leap forward in global sustainability efforts and demonstrate the great potential of ML to combat climate change.

2026-10-12

arXiv (preprint)

Mixture of GFlowNets

Chirayu Anant Haryan

MD Vasimuddin

Sanchit Misra

Generative Flow Networks (GFlowNets) learn a policy that samples compositional objects in proportion to an unnormalized reward …

2026-10-11

arXiv (preprint)

Multi-Objective Amortized CDRH3 Sequence Sampling Diversifies Promising Antibody Candidates

Joseph Viviano

Almer Van Der Sloot

Quentin Fournier

Robert M. Vernon

Christopher James Langmead

Antibody therapeutics have transformed treatment across diverse therapeutic areas and have become central to modern medicine. While recent m… (see more)achine learning approaches aim to accelerate antibody discovery through structure-based design and target-affinity optimization, successful candidates must also satisfy broad developability constraints, including structural compatibility, low immunogenicity, and favorable biophysical properties, while remaining sufficiently diverse to provide sufficient coverage of design space to mitigate later-stage development failures. We introduce a preference-conditioned multi-objective generative flow network (GFN) fine-tuning pipeline for fixed-framework CDRH3 loop redesign that fine-tunes a biological sequence model against many inexpensive proxies covering backbone compatibility, sequence likelihood, immunogenicity risk, novelty, and developability liabilities. It combines up to fourteen predictive rewards in a single policy trained to sample CDRH3 loops across 623 SAbDab antibody structures, leveraging a reward reduction strategy to preserve Pareto dominance under this learned map. We also introduce a novel autoregressive span-in-filling variant of AMPLIFY-350M that serves both as a scaffold-conditioned CDRH3 generation prior and as a direct conditional likelihood scorer. We evaluated reward scores, diversity, and design quality on loops sampled on 45 held-out antibody complexes to assess structural consistency, confidence, aggregation propensity, stability, and antigen binding. Compared with the inverse-folding methods ProteinMPNN and ESM-IF, our approach produces roughly 10x more distinct acceptable candidates per design at 50% sequence identity, while holding 10% of the 14-objective Pareto front pooled across methods, trading per-sample quality for greater exploration of novel possible structures.

2026-10-11

arXiv (preprint)

How to Verify Probabilistic Consistency of Predictive Models

Orr Paradise

Oliver Richardson

Shafi Goldwasser

When a probabilistic predictor answers many conditional-probability queries, are its answers self-consistent, and can this be verified in po… (see more)lynomial time? This problem is of interest for AI safety, where safety is derived from honesty about probabilistic predictions of unwanted outcomes potentially caused by an AI action. We construct an interactive PCP as follows. Let a predictive model be specified by a probability circuit P and a circuit Q which outputs confidence in predictions. Together, P and Q implicitly specify exponentially many probabilistic claims. We show a protocol in which a polynomial-time verifier can verify the approximate consistency of (P,Q). The verifier is given the pair of circuits (P,Q), which it evaluates at only a few points; alongside them it is given a proof oracle, an encoding of a witnessing probability distribution allegedly consistent with the predictions of (P,Q), which it reads at a few locations while interacting with a single untrusted prover. En route, we must ensure the existence of a sparse witnessing distribution consistent with the model's predictions. To do so, we first consider witness distributions for the consistency of explicit probabilistic claims, rather than claims specified by a predictor: say m claims, each of the form Pr[Y = 1 | X = x] = p, over n Boolean variables. Building on work initiated by Nilsson (Artif. Intell., 1986), we place l_2-approximate probabilistic consistency of explicit claims in NP, with certificates of length O(mn + log B) in the input bit-precision B; we further show how a small additive completeness-soundness gap removes the dependence on B. Together these results provide a complexity-theoretic foundation for certifying the self-consistency of probabilistic predictors. We view our interactive PCP as a first step toward training predictive models to prove their own consistency.

2026-08-10

arXiv (preprint)

Towards Trustworthy Physical AI: From Theory to Practice Across Life Cycle

Wang Yang

Hongxuan Liu

Xinghui Xu

Arjun Menon

Xiaoran Cai

Yunyu He

Alex Tarvo

Jingzong Zhou

Mengzhong Ma

Xinpeng Wei

Yi Yu

Shaobo Wang

Cheng Peng

Aoran Jiao

Alexei Korolev

Yanyan Zhang

Kai Ye

Xinpeng Li

Chengquan Guo

Jingjing Fu … (see 23 more)

Nicholas Bai

Yongjun He

Junru Ren

Silei Ren

Mohamad Louai Shehab

Keshu Cai

Nathaniel Dennler

Traian Tus

Gaoyue Zhou

Abdullah Garra

Mason Nakamura

George Ortiz

Marius Urbonas

Lars Johannsmeier

Rohit Sharma

Suraj Deshmukh

Michael Spranger

Vinayak Gupta

Xu Chen

Ashis G. Banerjee

Yuxiang Feng

Peng Qi

Physical AI refers to AI systems that understand, reason about, and act in accordance with the physical world and its underlying laws, dynam… (see more)ics, and constraints. Unlike conventional AI systems, physical AI interacts continuously with uncertain physical environments, and its actions produce consequences that are physically irreversible. As existing trustworthy AI frameworks have been developed primarily for digital AI systems, they do not fully capture the distinctive challenges of physical AI, such as physical safety, cyber-physical security, and physical manufacturing process. To address this gap, we present a survey of trustworthy physical AI principles. First, we characterize the core capabilities and challenges of physical AI. Second, we examine the role of physics in AI. Third, we trace the end-to-end physical AI life cycle across five core stages and introduce Trustworthy Physical AI Operationalization (T-PAIO). Fourth, we develop the Trustworthy Physical AI (T-PAI) framework, a theoretical framework that organizes key trustworthiness principles and provides a foundation for governing trustworthy physical AI systems.

2026-07-23

arXiv (preprint)

Insuring AI: Incentivising Safe and Secure Deployment

Agni Orfanoudaki

Carsten Maple

Matthew Wicker

Kwok-Yan Lam

Marcin Detyniecki

Lukasz Szpruch

2026-07-19

Harvard Data Science Review (published)

Realigning AI technology towards the Sustainable Development Goals

Rundong Feng

Ziming Li

Bojie Fu

Huadong Guo

Jianguo Liu

2026-07-13

Nature Machine Intelligence (published)