Yann Pequignot

Machine Learning publications

Interpretability

How to provide information on models’ decisions that is both trustworthy and useful?

GRANITE: A Generalized Regional Framework for Identifying Agreement in Feature-Based Explanations

J Herbinger, G Laberge, M Muschalik, Y Pequignot, MN Wright, F Fumagalli.

AISTATS 2026.

Tackling the XAI Disagreement Problem with Regional Explanations

G Laberge, Y Pequignot, M Marchand, F Khomh.

AISTATS 2024.

Partial order in chaos: consensus on feature attributions in the rashomon set

G Laberge, Y Pequignot, A Mathieu, F Khomh, M Marchand.

JMLR, 2023.

Understanding Interventional TreeSHAP: How and Why it Works

G Laberge, Y Pequignot. ArXiv, 2022.

Safe learning from interactions

Learning through interactions is an extremely powerful approach, yet it also carries inherent risks. It is therefore essential to develop theories and methods to enable agents to learn and explore safely.

SHAPO: Sharpness-Aware Policy Optimization for Safe Exploration

K Mani, Y Pequignot, V Mai, L Paull.

ICLR 2026.

Robustness

Training data and performance metrics only partially specify the task. Still, ML models are expected to perform reliably in diverse situations. What’s the gap between the training objective and the intendend behavior of a model?

Robust Fine-Tuning from Non-Robust Pretrained Models: Mitigating Suboptimal Transfer With Adversarial Scheduling

J Ngnawé, M Heuillet, S Sahoo, Y Pequignot, O Ahmad, A Durand, F Precioso, C Gagné.

ICLR 2026.

GROOD: Gradient-Aware Out-of-Distribution Detection

M ElAraby, S Sahoo, Y Pequignot, P Novello, L Paull.

TMLR 2025.

A Guide to Robust Generalization: The Impact of Architecture, Pre-training, and Optimization Strategy

M Heuillet, R Bhagwatkar, J Ngnawé, Y Pequignot, A Larouche, C Gagné, I Rish, O Ahmad, A Durand.

Workshop @ NeurIPS 2025: Reliable ML from Unreliable Data.

A Layer Selection Approach to Test Time Adaptation

S Sahoo, M ElAraby, J Ngnawe, Y Pequignot, F Precioso, C Gagné.

AAAI 2025.

Detecting Brittle Decisions for Free: Leveraging Margin Consistency in Deep Robust Classifiers

J Ngnawé, S Sahoo, Y Pequignot, F Precioso, C Gagné.

Neurips 2024.

TrackPGD: Efficient Adversarial Attack using Object Binary Masks against Robust Transformer Trackers

FN Nokabadi, Y Pequignot, JF Lalonde, C Gagné.

Workshop @ NeurIPS 2024: AdvML-Frontiers.

How to certify machine learning based safety-critical systems? A systematic literature review

F Tambon, G Laberge, L An, A Nikanjam, PSN Mindom, Y Pequignot, F Khomh, G Antoniol, E Merlo, F Laviolette.

Automated Software Engineering, 2022

Out-of-distribution detection for regression tasks: parameter versus predictor entropy

Y Pequignot, M Alain, P Dallaire, A Yeganehparast, P Germain, J Desharnais, F Laviolette. 2021, arXiv Preprint.

Mathematical publications

Researchers in all areas have imagined a myriad of abstract objects, many of them infinite, and inevitably followed by an infinite suite. What does it mean to understand them? How does a mathematician venture to make sense of these infinities he has imagined? Perhaps, one attempt could be to organise them, to arrange them, to order them.

Ph.D. Thesis

Selected undergraduate works