Germain Vivier-Ardisson

Welcome to my personal website !

I am a second year PhD student under the supervision of Axel Parmentier (CERMICS, Ecole des Ponts) and Mathieu Blondel (Google DeepMind). My research focuses on interfacing combinatorial optimization and ML methods through regularization.

Before that, I graduated from Ecole polytechnique (X2020) and got a master’s degree from Sorbonne Université in Learning and Algorithms (Master M2A).

news

Jul 31, 2026	Our paper Learning with Local Search MCMC Layers has been accepted to TMLR with a Featured Certification !
Jun 03, 2026	New pre-print out ! Regularized Large Neighborhood Search
Apr 30, 2026	Our paper Autoregressive Language Models are Secretly Energy-Based Models: Insights into the Lookahead Capabilities of Next-Token Prediction has been accepted to ICML 2026 !
Jan 31, 2026	New pre-print out ! Differentiable Knapsack and Top-k Operators via Dynamic Programming
Dec 07, 2025	I gave a 15min talk at the DiffCoALG workshop @ NeurIPS 2025 to present our Learning with Local Search MCMC Layers paper !

selected publications

Regularized Large Neighborhood Search

Germain Vivier-Ardisson, Laurent Demonet , Axel Parmentier , and Mathieu Blondel

arXiv preprint arXiv:2606.02294, 2026

Bib

@article{vivierardisson2026regularizedlargeneighborhoodsearch,
  title = {Regularized Large Neighborhood Search},
  author = {Vivier-Ardisson, Germain and Demonet, Laurent and Parmentier, Axel and Blondel, Mathieu},
  journal = {arXiv preprint arXiv:2606.02294},
  year = {2026},
  url = {https://arxiv.org/abs/2606.02294},
}

Differentiable Knapsack and Top-k Operators via Dynamic Programming

Germain Vivier-Ardisson, Michaël E. Sander , Axel Parmentier , and Mathieu Blondel

arXiv preprint arXiv:2601.21775, 2026

Bib

@article{vivier2026differentiable,
  title = {Differentiable Knapsack and Top-k Operators via Dynamic Programming},
  author = {Vivier-Ardisson, Germain and Sander, Micha{\"e}l E. and Parmentier, Axel and Blondel, Mathieu},
  journal = {arXiv preprint arXiv:2601.21775},
  year = {2026},
  url = {https://arxiv.org/abs/2601.21775},
}

Autoregressive Language Models are Secretly Energy-Based Models: Insights into the Lookahead Capabilities of Next-Token Prediction

Mathieu Blondel , Michael E. Sander , Germain Vivier-Ardisson, Tianlin Liu , and 1 more author

In , 2026

Bib

@inproceedings{blondel2026autoregressive,
  title = {Autoregressive Language Models are Secretly Energy-Based Models: Insights into the Lookahead Capabilities of Next-Token Prediction},
  author = {Blondel, Mathieu and Sander, Michael E. and Vivier-Ardisson, Germain and Liu, Tianlin and Roulet, Vincent},
  journal = {ICML},
  year = {2026},
  url = {https://arxiv.org/abs/2512.15605},
}

Learning with Local Search MCMC Layers

Germain Vivier-Ardisson, Mathieu Blondel , and Axel Parmentier

TMLR (Featured Certification), 2026

Abs Bib

Integrating combinatorial optimization layers into neural networks has recently attracted significant research interest. However, many existing approaches lack theoretical guarantees or fail to perform adequately when relying on inexact solvers. This is a critical limitation, as many operations research problems are NP-hard, often necessitating the use of neighborhood-based local search heuristics. These heuristics iteratively generate and evaluate candidate solutions based on an acceptance rule. In this paper, we introduce a theoretically-principled approach for learning with such inexact combinatorial solvers. Inspired by the connection between simulated annealing and Metropolis-Hastings, we propose to transform problem-specific neighborhood systems used in local search heuristics into proposal distributions, implementing MCMC on the combinatorial space of feasible solutions. This allows us to construct differentiable combinatorial layers and associated loss functions. Replacing an exact solver by a local search strongly reduces the computational burden of learning on many applications. We demonstrate our approach on a large-scale dynamic vehicle routing problem with time windows.
@article{vivier2025learning, title = {Learning with Local Search MCMC Layers}, author = {Vivier-Ardisson, Germain and Blondel, Mathieu and Parmentier, Axel}, journal = {TMLR (Featured Certification)}, year = {2026}, url = {https://arxiv.org/abs/2505.14240} }

CF-OPT: Counterfactual Explanations for Structured Prediction

Germain Vivier-Ardisson, Alexandre Forel , Axel Parmentier , and Thibaut Vidal

ICML, 2024

Bib Code

@article{vivier--ardisson_cf-opt_2024,
  title = {{CF}-{OPT}: {Counterfactual} {Explanations} for {Structured} {Prediction}},
  shorttitle = {{CF}-{OPT}},
  url = {http://arxiv.org/abs/2405.18293},
  author = {Vivier-Ardisson, Germain and Forel, Alexandre and Parmentier, Axel and Vidal, Thibaut},
  year = {2024},
  journal = {ICML},
}