I am currently an AI research scientist at Thales cortAIx-Labs, in Paris region. My research interests span explainable machine learning (XAI), causal inference, tree ensembles, decision aid, deep learning, uncertainty quantification, and active learning. I am also driven by practical applications and software development.

I have ten years of experience as a machine learning researcher and data scientist in aeronautic/aerospace and tech industries, with also a strong interest in healthcare applications. I hold a PhD in machine learning and mathematical statistics, on the topic of explainable and interpretable ML (XAI). In my research, I address both the design of new algorithms, and the theoretical analysis of their mathematical properties.

My resume is available here.

Research interests

• Explainable Machine Learning - XAI (functional decompositions, variable importance, rule learning)
• Causal Inference (heterogeneous treatment effects, causal forests, causal discovery)
• Tree Ensembles (random forests, boosted trees)
• Decision Aid (MCDA, Neur-HCI)
• Deep Learning (CNN, VLM, GNN)
• Uncertainty Quantification (Bayesian inference, conformal predictions)
• Active Learning (design of experiments, Bayesian optimization)

Publications

Preprints

Published/Accepted papers

• C. Bénard. Tree Ensemble Explainability through the Hoeffding Functional Decomposition and TreeHFD Algorithm. In Advances in Neural Information Processing Systems 38 (NeurIPS 2025), in press, 2025.
• C. Bénard, J. Josse. Variable importance for causal forests: breaking down the heterogeneity of treatment effects. Accepted to Journal of Causal Inference, 2025.
• N. Honoré, C. Bénard, and C. Labreuche. Sensitivity Analysis in Surveillance Performance Monitoring for Air Traffic Management. 2025 28th International Conference on Information Fusion (FUSION), IEEE, 2025.
• C. Bénard, J. Näf, and J. Josse. MMD-based Variable Importance for Distributional Random Forest. Proceedings of The 27th International Conference on Artificial Intelligence and Statistics (AISTATS 2024), PMLR 238:1324-1332, 2024.
• C. Bénard, B. Staber, and S. Da Veiga. Kernel Stein Discrepancy thinning: a theoretical perspective of pathologies and a practical fix with regularization. Advances in Neural Information Processing Systems 36 (NeurIPS 2023), 2023.
• C. Bénard, S. Da Veiga, and E. Scornet. Mean decrease accuracy for random forests: inconsistency, and a practical solution via the Sobol-MDA. Biometrika, 109:881-900, 2022.
• C. Bénard, G. Biau, S. Da Veiga, and E. Scornet. SHAFF: Fast and consistent SHApley eFfect estimates via random Forests. In Proceedings of The 25th International Conference on Artificial Intelligence and Statistics (AISTATS 2022), PMLR 151:5563-5582, 2022.
• C. Bénard, G. Biau, S. Da Veiga, and E. Scornet. Interpretable random forests via rule extraction. In Proceedings of The 24th International Conference on Artificial Intelligence and Statistics (AISTATS 2021), PMLR 130:937–945, 2021.
• C. Bénard, G. Biau, S. Da Veiga, and E. Scornet. SIRUS: Stable and Interpretable RUle Set for classification. Electronic Journal of Statistics, 15:427–505, 2021.

Book

• B. Iooss, R. Kenett, P. Secchi, B.M. Colosimo, F. Centofanti, C. Bénard, S. Da Veiga, E. Scornet, S. N. Wood, Y. Goude, M. Fasiolo. Interpretability for Industry 4.0: Statistical and Machine Learning Approaches, Editors: A. Lepore, B. Palumbo, J.-M. Poggi, Springer 2022.

Academic publications

• PhD thesis Random forests and interpretability of learning algorithms, 6 December 2021. Supervisors : Gérard Biau, Erwan Scornet, and Sébastien Da Veiga.

Software

• treehfd: A python module to explain xgboost models with the Hoeffding functional decomposition.
• Kernax: Kernax is a package that implements KSD-based algorithms for post-processing MCMC outputs. It is based on JAX and works on CPU as well as GPU.
• Lagun: platform providing a user-friendly interface to methods and algorithms dedicated to the exploration of numerical simulations and the analysis of datasets (design of experiments, conditional gaussian processes, sensitivity analysis, uncertainty propagation, optimization).
• sirus (Stable and Interpretable RUle Set): a regression and classification algorithm based on random forests, which takes the form of a short list of rules. SIRUS combines the simplicity of decision trees with a predictivity close to random forests.
• shaff (SHApley eFfects via random Forests): a fast and accurate algorithm to estimate Shapley effects.
• sobolMDA: the Sobol-MDA is a variable importance measure for random forests, fixing the flaws of Breiman’s MDA.
• vimp-causal-forests: a variable importance measure for causal forests.

Talks

• October 2025, MobiliT.AI 2025, Toulouse, France.
• March 2024, SIAM UQ 2024, Trieste, Italy.
• December 2023, NeurIPS@Paris, Sorbonne Université, Paris.
• December 2023, 2nd Nice Workshop on Interpretability, Laboratoire J.A. Dieudonné, Université Côte d’Azur, Nice.
• August 2023, EcoSta 2023, Tokyo, Japan.
• March 2023, CSMA Junior, Paris.
• December 2022, CMStatistics 2022, King’s College, London, UK.
• March 2022, AISTATS 2022, virtual conference.
• March 2022, Séminaire de Probabilités et Statistiques, Laboratoire de Mathématiques d’Orsay, Université Paris-Saclay, Orsay.
• January 2022, Séminaire de Probabilités et Statistiques, Laboratoire J.A. Dieudonné, Université Côte d’Azur, Nice.
• September 2021, École Thématique sur les Incertitudes en Calcul Scientifique (ETICS2021), Erdeven, France.
• July 2021, ENBIS Workshop: Interpretability of Industry 4.0, Naples, Italy.
• June 2021, 52èmes Journées de Statistiques 2021, Nice, France.
• April 2021, GDR Mascot-Num, virtual conference.
• April 2021, SIMPAS Group meeting, Centre de Mathématiques Appliquées, Ecole Polytechnique, Saclay, France.
• April 2021, AISTATS 2021, virtual conference.
• June 2019, 51èmes Journées de Statistiques 2019, Nancy, France.

Teaching

• Instructor in statistics and optimization for mechanical engineers, 3-day sessions multiple times a year since 2018 (machine learning, uncertainty quantification, design of experiments, constrained optimization…)