Portrait de Irina Rish

Irina Rish

Membre académique principal
Chaire en IA Canada-CIFAR
Professeure titulaire, Université de Montréal
Sujets de recherche
Apprentissage en ligne
Apprentissage multimodal
Apprentissage par renforcement
Apprentissage profond
Modèles génératifs
Neurosciences computationnelles
Traitement du langage naturel

Biographie

Irina Rish est professeure titulaire à l'Université de Montréal (UdeM), où elle dirige le Laboratoire d'IA autonome. Membre du corps professoral de Mila – Institut québécois d’intelligence artificielle, elle est titulaire d'une chaire d'excellence en recherche du Canada (CERC) et d'une chaire en IA Canada-CIFAR. Irina dirige le projet INCITE du ministère américain de l'Environnement au sujet des modèles de fondation évolutifs sur les superordinateurs Summit et Frontier à l'Oak Ridge Leadership Computing Facility (OLCF). Elle est cofondatrice et directrice scientifique de Nolano.ai.

Ses recherches actuelles portent sur les lois de mise à l'échelle neuronale et les comportements émergents (capacités et alignement) dans les modèles de fondation, ainsi que sur l'apprentissage continu, la généralisation hors distribution et la robustesse. Avant de se joindre à l'UdeM en 2019, Irina était chercheuse au Centre de recherche IBM Thomas J. Watson, où elle a travaillé sur divers projets à l'intersection des neurosciences et de l'IA, et dirigé le défi NeuroAI. Elle a reçu plusieurs prix IBM : ceux de l’excellence et de l’innovation exceptionnelle (2018), celui de la réalisation technique exceptionnelle (2017), et celui de l’accomplissement en recherche (2009). Elle détient 64 brevets et a écrit plus de 120 articles de recherche, plusieurs chapitres de livres, trois livres publiés et une monographie sur la modélisation éparse.

Étudiants actuels

Doctorat - Concordia
Superviseur⋅e principal⋅e :
Maîtrise recherche - Concordia
Superviseur⋅e principal⋅e :
Visiteur de recherche indépendant - -
Collaborateur·rice de recherche - UdeM
Maîtrise recherche - UdeM
Doctorat - UdeM
Superviseur⋅e principal⋅e :
Collaborateur·rice de recherche - UdeM
Collaborateur·rice alumni - UdeM
Doctorat - Concordia
Superviseur⋅e principal⋅e :
Collaborateur·rice de recherche
Collaborateur·rice de recherche - UdeM
Collaborateur·rice de recherche - McGill
Superviseur⋅e principal⋅e :
Doctorat - UdeM
Co-superviseur⋅e :
Collaborateur·rice de recherche - Polytechnique
Doctorat - McGill
Superviseur⋅e principal⋅e :
Maîtrise recherche - UdeM
Co-superviseur⋅e :
Doctorat - McGill
Superviseur⋅e principal⋅e :
Collaborateur·rice de recherche
Doctorat - Concordia
Superviseur⋅e principal⋅e :
Collaborateur·rice de recherche - UdeM
Baccalauréat - McGill
Maîtrise recherche - UdeM
Doctorat - McGill

Publications

Emergent Reasoning via Recursive Latent Reinforcement Pretraining
Large language models (LLMs) often rely on explicit chain-of-thought (CoT) traces to solve multi-step reasoning problems, but these traces i… (voir plus)ncrease inference cost, expose brittle prompt dependence, and complicate training objectives. We study an alternative: \emph{latent deliberation} implemented as a small recurrent refinement module that performs multiple internal ``thinking`` steps while keeping the external sequence length fixed. We introduce \textbf{Recursive Latent Reinforcement Pretraining (RLRP)}, a training recipe that augments a base causal LLM with a shared latent head executed for
AIF-GEN: Open-Source Platform and Synthetic Dataset Suite for Reinforcement Learning on Large Language Models
On the Adversarial Robustness of Discrete Image Tokenizers
Nicolas Flammarion
Francesco Croce
Discrete image tokenizers encode visual inputs as sequences of tokens from a finite vocabulary and are gaining popularity in multimodal syst… (voir plus)ems, including encoder-only, encoder-decoder, and decoder-only models. However, unlike CLIP encoders, their vulnerability to adversarial attacks has not been explored. Ours being the first work studying this topic, we first formulate attacks that aim to perturb the features extracted by discrete tokenizers, and thus change the extracted tokens. These attacks are computationally efficient, application-agnostic, and effective across classification, multimodal retrieval, and captioning tasks. Second, to defend against this vulnerability, inspired by recent work on robust CLIP encoders, we fine-tune popular tokenizers with unsupervised adversarial training, keeping all other components frozen. While unsupervised and task-agnostic, our approach significantly improves robustness to both unsupervised and end-to-end supervised attacks and generalizes well to unseen tasks and data. Unlike supervised adversarial training, our approach can leverage unlabeled images, making it more versatile. Overall, our work highlights the critical role of tokenizer robustness in downstream tasks and presents an important step in the development of safe multimodal foundation models.
$\mu$LO: Compute-Efficient Meta-Generalization of Learned Optimizers
Learned optimizers (LOs) have the potential to significantly reduce the wall-clock training time of neural networks. However, they can strug… (voir plus)gle to optimize unseen tasks (*meta-generalize*), especially when training networks wider than those seen during meta-training. To address this, we derive the Maximal Update Parametrization (
Deep neural networks divide and conquer dihedral multiplication
We find multilayer perceptrons and transformers both universally learn an instantiation of the same divide-and-conquer algorithm that requir… (voir plus)es only a logarithmic number of neural representations to solve dihedral multiplication. Clustering neurons based on similar activation behaviour reveals remarkably clear structure: each neural representation corresponds to a Cayley graph. To our knowledge, this is the first work that fully characterizes and describes all neural representations that are learnable on a dataset, while prior work on group multiplications studied neuron-level behavior, or preliminarily investigated cluster behavior. Thus, we can understand the algorithm networks universally learn at three levels of abstraction: 1) Neurons activate on coset or approximate coset structure of the dihedral group. 2) Groups of neurons together form neural representations that act to divide the dataset into different subproblems, being Cayley graphs, where the equivalence class of the answer is computed. 3) The global algorithm then linearly combines each neural representation (subproblem) together at the logits. This work provides a deep case study and provides the community with a very well understood toy model for interpretability, as well as makes steps toward proving the conjecture that DNNs will divide and conquer all group multiplication tasks.
MEEGNet: An open source python library for the application of convolutional neural networks to MEG
Annalisa Pascarella
Yann Harel
Dehgan Arthur
Rish Irina
Jerbi Karim
Abstract Artificial Neural Networks (ANNs) are rapidly gaining traction in neuroscience, proving invaluable for decoding and modeling brain … (voir plus)signals from techniques such as electroencephalography (EEG) and functional magnetic resonance imaging (fMRI). Although these networks are beginning to find applications in magnetoencephalography (MEG), their use in this domain is still in the early stages. Here, we introduces MEEGNet, a novel Python library paired with an intuitive convolutional neural network (CNN) architecture designed primarily for MEG data, yet adaptable to EEG signals. The MEEGNet model was trained and cross-validated using MEG data from 643 participants across four classification tasks, including auditory and visual stimulus classification and age prediction. Our model achieves competitive performance across all tasks, with a notable balance of accuracy and efficiency—for instance, reaching 92.70% test accuracy in an auditory vs. visual classification task while maintaining shorter training times than other architectures. The MEEGNet pipeline also integrates latent space visualization tools, adapted for MEG and EEG data. These include saliency maps and Grad-CAM methods, which enhance the interpretability of ANN-based classification and help address the black-box critique of such models. Importantly, the MEEGNet library is designed for extensibility, allowing the neuroscience and machine learning communities to add functionalities and ANN models. By prioritizing usability, transparency, and interpretability, MEEGNet empowers MEG- and EEG-based research with a user-friendly and modular framework.
Position: Collusion Risks Among AI Reasoning Agents Justify Certification Requirements for Making Market Decisions
This position paper argues that AI agents with chain-of-thought reasoning capabilities are predisposed to exhibit collusive behavior and sho… (voir plus)uld be required to obtain behavioral certification before making decisions that affect economic markets. This is because integrating these agents into society could collapse the legal evidentiary distinction between competition and collusion among independent firms without eroding the economic harm distinction. Experiments with DeepSeek-R1 agents in the Bertrand oligopoly pricing domain reveal a tendency towards tacit collusion that persists even when humans prompt the agents not to collude. We further show that the chain-of- thought of these agents can be steered toward either extremely collusive or highly competitive behavior in a way that is not semantically detectable by another LLM analyzing the reasoning traces. As a result, deploying reasoning agents for market decisions leads to collusive economic outcomes without any evidence of conspiracy or intent. Thus, certification based on observed behavior in representative situations is necessary to prevent collusion. We provide preliminary evidence that such agents can be steered in a generalizable way toward efficient competitive equilibria. However, developing a comprehensive behavioral certification will be required before these models can be deployed in real-world markets while ensuring their stability and efficiency.
RobustGenBench: A Benchmark for Robust Generalization to Adversarial and Common Perturbations, with Applications to Vision and Vision-Enabled Large Language Models
Handling Delay in Real-Time Reinforcement Learning
Real-time reinforcement learning (RL) introduces several challenges. First, policies are constrained to a fixed number of actions per second… (voir plus) due to hardware limitations. Second, the environment may change while the network is still computing an action, leading to observational delay. The first issue can partly be addressed with pipelining, leading to higher throughput and potentially better policies. However, the second issue remains: if each neuron operates in parallel with an execution time of
Maxwell's Demon at Work: Efficient Pruning by Leveraging Saturation of Neurons
When training neural networks, dying neurons -- units becoming inactive or saturated -- are traditionally seen as harmful. This paper sheds … (voir plus)new light on this phenomenon. By exploring the impact of various hyperparameter configurations on dying neurons during training, we gather insights on how to improve upon sparse training approaches to pruning. We introduce Demon Pruning (DemP), a method that controls the proliferation of dead neurons through a combination of noise injection on active units and a one-cycle schedule regularization strategy, dynamically leading to network sparsity. Experiments on CIFAR-10 and ImageNet datasets demonstrate that DemP outperforms existing dense-to-sparse structured pruning methods, achieving better accuracy-sparsity tradeoffs and accelerating training by up to 3.56
Critical Role of EEG Signals in Assessment of Sex-Specific Insights in Neurological Diagnostics via Machine Learning Approach
Mohammad-Javad Darvishi-Bayazi
Mohammad Sajjad Ghaemi
Jocelyn Faubert
Abstract

Early detection and diagnosis of pathology are essential for efficient treatment and therapeutic … (voir plus)interventions. The emergence of Artificial Intelligence (AI) and deep machine learning techniques have demonstrated the promising capability of brain imaging data to predict various pathological diseases. However, plenty of diseases have imbalanced distribution across different sexes. Furthermore, the impact of sex-specific patterns and biomarkers in predicting diseases has remained unexplored as a fundamental subject matter to inform the treatment paradigms. This paper underscored the generalization and transferability of sex-related patterns in functional data, specifically Electroencephalogram (EEG) signals through Artificial Deep Neural Networks. We conducted training on a broad spectrum of EEG recordings involving participants ranging from 221 to 12,000, including healthy and pathological subjects. Our evaluation leveraged datasets from various sources and participant groups, featuring distribution shifts. While the artificial models demonstrated accurate sex detection on datasets without fine-tuning, their performance declined with significant distribution shifts. Furthermore, we explored the relationship between sex and pathology by visualizing salient features for target detection in distinct subgroups. Our findings revealed unprecedented insights into the negligible role of sex-specific patterns in pathology detection despite the presence of prominent and consistent patterns within sex groups. These results are essential for developing more robust and unbiased AI models for disease prediction and informing the treatment paradigms.

CAVE: Detecting and Explaining Commonsense Anomalies in Visual Environments
Syrielle Montariol
Angelika Romanou
Beatriz Borges
Antoine Bosselut