Doina Precup

Samin Yeasar Arnob

PhD - McGill University

arnobsam@mila.quebec

sumana.basu@mail.mcgill.ca

Sumana Basu

Collaborating Alumni - McGill University

Co-supervisor :

Adriana Romero Soriano

Master's Research - McGill University

gunbir-singh.baveja@mila.quebec

Collaborating Alumni - McGill University

cheluver@mila.quebec

Raymond Chua

PhD - McGill University

Co-supervisor :

PhD - McGill University

Principal supervisor :

David Meger

chungwes@mila.quebec

Jonathan Colaço Carr

Master's Research - McGill University

Principal supervisor :

Prakash Panangaden

jonathan.colaco-carr@mila.quebec

Élodie Coté-Gauthier

Master's Research - McGill University

Co-supervisor :

Isabeau Prémont-Schwarz

elodie.cote-gauthier@mila.quebec

Franco Del Balso

Master's Research - McGill University

franco.delbalso@mila.quebec

busra-tugce.gurbuz@mila.quebec

Jesse Farebrother

PhD - McGill University

Principal supervisor :

Marc Gendron-Bellemare

PhD - McGill University

Principal supervisor :

Eilif B. Muller

Mohammad Sami Nur Islam Islam

Jonathan Hu

Master's Research - McGill University

jonathan.hu@mila.quebec

Howard Huang

PhD - McGill University

huanghow@mila.quebec

Master's Research - McGill University

mohammad-sami-nur.islam@mila.quebec

Hangzhan Jin

PhD - Polytechnique Montréal

jinhangzhan@gmail.com

flemming.kondrup@mila.quebec

Flemming Kondrup

Postdoctorate - McGill University

Jonathan Lebensold

Collaborating Alumni - McGill University

Collaborating Alumni - McGill University

jeremy.lucas@mila.quebec

Ray Luo

PhD - McGill University

Principal supervisor :

Xujie Si

luo.ziyan@mila.quebec

Sicheng Lyu

Master's Research - McGill University

sicheng.lyu@mila.quebec

G McCracken

PhD - McGill University

mccrackg@mila.quebec

shahrad.mohammadzadeh@mila.quebec

Shahrad Mohammadzadeh

Master's Research - McGill University

Principal supervisor :

Reihaneh Rabbany

padideh.nouri@mila.quebec

Collaborating researcher - McGill University

Co-supervisor :

Irina Rish

moisescg@mila.quebec

Padideh Nouri

PhD - Université de Montréal

PhD - McGill University

Co-supervisor :

Nate Rahn

PhD - McGill University

Principal supervisor :

Marc Gendron-Bellemare

nathan.rahn@mila.quebec

Mandana Samiei

PhD - McGill University

Co-supervisor :

Blake Richards

samiemandana@gmail.com

PhD - McGill University

Co-supervisor :

Paul Masset

manoosh.samiei@mila.quebec

PhD - McGill University

Nishanth Anand Vemgal

PhD - McGill University

PhD - McGill University

Co-supervisor :

Samira Ebrahimi Kahou

Research Intern - McGill University

guangyuan.wang@mila.quebec

Zihan Wang

PhD - McGill University

zihan.wang@mila.quebec

william.potter@mila.quebec

Steve Wen

Master's Research - McGill University

Co-supervisor :

Gregory Dudek

steve.wen@mila.quebec

Will William

Research Intern - McGill University

shuyuan.zhang@mila.quebec

Zijing Wu

PhD - McGill University

Principal supervisor :

Paul Masset

zijing.wu@mila.quebec

PhD - McGill University

Harry Zhao

Collaborating Alumni - McGill University

Co-supervisor :

Blog Posts

Generic thumbnail for Mila Blog articles.

February 22, 2024

Skipper: Combining Spatial and Temporal Abstraction for Better Generalization

Mingde Harry Zhao

Safa Alver

Harm van Seijen

Romain Laroche

Doina Precup

Yoshua Bengio

Read the article

Publications

Plasticity as the Mirror of Empowerment

David Abel

Michael Bowling

Andre Barreto

Will Dabney

Shi Dong

Steven Hansen

Anna Harutyunyan

Khimya Khetarpal

Clare Lyle

Razvan Pascanu

Georgios Piliouras

Jonathan Richens

Mark Rowland

Tom Schaul

Satinder Singh

Agents are minimally entities that are influenced by their past observations and act to influence future observations. This latter capacity … (see more)is captured by empowerment, which has served as a vital framing concept across artificial intelligence and cognitive science. This former capacity, however, is equally foundational: In what ways, and to what extent, can an agent be influenced by what it observes? In this paper, we ground this concept in a universal agent-centric measure that we refer to as plasticity, and reveal a fundamental connection to empowerment. Following a set of desiderata on a suitable definition, we define plasticity using a new information-theoretic quantity we call the generalized directed information. We show that this new quantity strictly generalizes the directed information introduced by Massey (1990) while preserving all of its desirable properties. Under this definition, we find that plasticity is well thought of as the mirror of empowerment: The two concepts are defined using the same measure, with only the direction of influence reversed. Our main result establishes a tension between the plasticity and empowerment of an agent, suggesting that agent design needs to be mindful of both characteristics. We explore the implications of these findings, and suggest that plasticity, empowerment, and their relationship are essential to understanding agency

2025-12-02

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

Uncovering a Universal Abstract Algorithm for Modular Addition in Neural Networks

Gavin McCracken

Vincent Létourneau

Jonathan Love

We propose a testable universality hypothesis, asserting that seemingly disparate neural network solutions observed in the simple task of mo… (see more)dular addition are unified under a common abstract algorithm. While prior work interpreted variations in neuron-level representations as evidence for distinct algorithms, we demonstrate - through multi-level analyses spanning neurons, neuron clusters, and entire networks - that multilayer perceptrons and transformers universally implement the abstract algorithm we call the approximate Chinese Remainder Theorem. Crucially, we introduce approximate cosets and show that neurons activate exclusively on them. Furthermore, our theory works for deep neural networks (DNNs). It predicts that universally learned solutions in DNNs with trainable embeddings or more than one hidden layer require only O(log n) features, a result we empirically confirm. This work thus provides the first theory-backed interpretation of multilayer networks solving modular addition. It advances generalizable interpretability and opens a testable universality hypothesis for group multiplication beyond modular addition.

2025-12-02

Neural Information Processing Systems (Accept (poster))

Excitatory-Inhibitory Dynamics in Adaptive Decision-Making

Veronica Chelu

2025-11-15

Lecture Notes in Computer Science (published)

The Geometry and Topology of Modular Addition Representations

Colin Daniels

Jonathan Love

The Clock and Pizza interpretations, associated with neural architectures differing in either uniform or learnable attention, were introduce… (see more)d to argue that different architectural designs can yield distinct circuits for modular addition. Applying geometric and topological analyses to learned representations, we show that this is not the case: Clock and Pizza circuits are topologically and geometrically equivalent and are thus equivalent representations.

2025-11-12

TAG-DS/2025/Conference (poster)

Incorporating Spatial Information into Goal-Conditioned Hierarchical Reinforcement Learning via Graph Representations

Shuyuan Zhang

Zihan Wang

Xiao-Wen Chang

The integration of graphs with Goal-conditioned Hierarchical Reinforcement Learning (GCHRL) has recently gained attention, as intermediate g… (see more)oals (subgoals) can be effectively sampled from graphs that naturally represent the overall task structure in most RL tasks. However, existing approaches typically rely on domain-specific knowledge to construct these graphs, limiting their applicability to new tasks. Other graph-based approaches create graphs dynamically during exploration but struggle to fully utilize them, because they have problems passing the information in the graphs to newly visited states. Additionally, current GCHRL methods face challenges such as sample inefficiency and poor subgoal representation. This paper proposes a solution to these issues by developing a graph encoder-decoder to evaluate unseen states. Our proposed method, Graph-Guided sub-Goal representation Generation RL (G4RL), can be incorporated into any existing GCHRL method when operating in environments with primarily symmetric and reversible transitions to enhance performance across this class of problems. We show that the graph encoder-decoder can be effectively implemented using a network trained on the state graph generated during exploration. Empirical results indicate that leveraging high and low-level intrinsic rewards from the graph encoder-decoder significantly enhances the performance of state-of-the-art GCHRL approaches with an extra small computational cost in dense and sparse reward environments.

2025-10-05

Transactions on Machine Learning Research (accepted)

Rejecting Hallucinated State Targets during Planning

Mingde Zhao

Tristan Sylvain

Romain Laroche

Yoshua Bengio

In planning processes of computational decision-making agents, generative or predictive models are often used as "generators" to propose "ta… (see more)rgets" representing sets of expected or desirable states. Unfortunately, learned models inevitably hallucinate infeasible targets that can cause delusional behaviors and safety concerns. We first investigate the kinds of infeasible targets that generators can hallucinate. Then, we devise a strategy to identify and reject infeasible targets by learning a target feasibility evaluator. To ensure that the evaluator is robust and non-delusional, we adopted a design choice combining off-policy compatible learning rule, distributional architecture, and data augmentation based on hindsight relabeling. Attaching to a planning agent, the designed evaluator learns by observing the agent's interactions with the environment and the targets produced by its generator, without the need to change the agent or its generator. Our controlled experiments show significant reductions in delusional behaviors and performance improvements for various kinds of existing agents.

2025-10-05

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

proceedings.mlr.press

Reward the Reward Designer: Making Reinforcement Learning Useful for Clinical Decision Making

Sumana Basu

Adriana Romero

2025-09-21

NeurIPS.cc/2025/Workshop/WiML (published)

Unifying Mechanistic Interpretations of Neural Networks Trained on Modular Addition

Gavin McCracken

Vincent Létourneau

Jonathan Love

2025-09-21

NeurIPS.cc/2025/Workshop/WiML (published)

Capturing Individual Human Preferences with Reward Features

Andre Barreto

Vincent Dumoulin

Yiran Mao

Nicolas Perez-Nieves

Mark Rowland

Bobak Shahriari

Yann Dauphin

Hugo Larochelle

Reinforcement learning from human feedback usually models preferences using a reward model that does not distinguish between people. We argu… (see more)e that this is unlikely to be a good design choice in contexts with high potential for disagreement, like in the training of large language models. We propose a method to specialise a reward model to a person or group of people. Our approach builds on the observation that individual preferences can be captured as a linear combination of a set of general reward features. We show how to learn such features and subsequently use them to quickly adapt the reward model to a specific individual, even if their preferences are not reflected in the training data. We present experiments with large language models comparing the proposed architecture with a non-adaptive reward model and also adaptive counterparts, including models that do in-context personalisation. Depending on how much disagreement there is in the training data, our model either significantly outperforms the baselines or matches their performance with a simpler architecture and more stable training.

2025-09-17

NeurIPS.cc/2025/Conference (poster)

RL Fine-Tuning Heals OOD Forgetting in SFT

Mohammad Hamdaqa

The two-stage fine-tuning paradigm of Supervised Fine-Tuning (SFT) followed by Reinforcement Learning (RL) has empirically shown better reas… (see more)oning performance than one-stage SFT for the post-training of Large Language Models (LLMs). However, the evolution and mechanism behind the synergy of SFT and RL are still under-explored and inconclusive. In our study, we find the well-known claim "SFT memorizes, RL generalizes" is over-simplified, and discover that: (1) OOD performance peaks at the early stage of SFT and then declines (OOD forgetting), the best SFT checkpoint cannot be captured by training/test loss; (2) the subsequent RL stage does not generate fundamentally better OOD capability, instead it plays an \textbf{OOD restoration} role, recovering the lost reasoning ability during SFT; (3) The recovery ability has boundaries, \ie{} \textbf{if SFT trains for too short or too long, RL cannot recover the lost OOD ability;} (4) To uncover the underlying mechanisms behind the forgetting and restoration process, we employ SVD analysis on parameter matrices, manually edit them, and observe their impacts on model performance. Unlike the common belief that the shift of model capacity mainly results from the changes of singular values, we find that they are actually quite stable throughout fine-tuning. Instead, the OOD behavior strongly correlates with the \textbf{rotation of singular vectors}. Our findings re-identify the roles of SFT and RL in the two-stage fine-tuning and discover the rotation of singular vectors as the key mechanism. %reversing the rotations induced by SFT, which shows recovery from forgetting, whereas imposing the SFT parameter directions onto a RL-tuned model results in performance degradation. Code is available at https://github.com/xiaodanguoguo/RL_Heals_SFT

2025-09-07

arXiv (Cornell University) (preprint)

arxiv.org

Relative Trajectory Balance is equivalent to Trust-PCL

2025-08-31

ArXiv (preprint)