Publications

To Retain or to Adapt? Generalizing Continual Learning
Giulia Lanzillotta
Mandana Samiei
Doina Precup
Razvan Pascanu
Claire Vernade
The Continual Learning (CL) literature has long been driven by the goal of mitigating catastrophic forgetting. This objective rests on a per… (see more)vasive, often unstated assumption: that a lifelong learner should approximate the Joint-Task Learning (JTL) solution and retain all previously acquired knowledge. We challenge this retention-centered premise, arguing that in non-stationary environments prioritizing retention can impede real-time adaptation. Shifting the focus to the Average Lifelong Error (ALE), we formalize CL as an online optimization problem governed by the interaction between environmental and learning dynamics. We introduce Transfer Efficiency as a quantitative measure of the tension between Instability, the bias inherited from conflicting past experience, and Transient Error, the optimization cost of learning new tasks from scratch. Under mild convergence conditions, holding across linear and neural network models, this decomposition yields a Critical Task Duration: a closed-form threshold beyond which historical knowledge transitions from a warm-start advantage to an optimization liability whenever retention induces a positive stationary bias. We validate these theoretical predictions on continual image classification and reinforcement learning benchmarks. Finally, by connecting continual learning to the online learning framework of predictable sequences, we show that JTL is only one instance of a broader family of objectives, and we propose a new general class of continual learning algorithms, which we call Predictive Continual Learning. Predictive CL algorithms optimize expected future performance under an explicit, dynamically updated model of future tasks. As a proof of concept, we analyze a Window algorithm that interpolates between JTL and Independent-Task Learning (ITL), outperforming both under controlled distributional drift.
2026-07-05
arXiv (preprint)
doi.org
arxiv.org
Contrasting invasion potential of Ciona robusta and C. intestinalis in Subantarctic and Antarctic ecosystems under warming scenarios: insights of ecological niche models and physiological data
Zambra López-Farrán
Luis Enrique Angeles-Gonzalez
Alejandro Ortiz
Fernando Diaz
Jorge M. Navarro
2026-07-04
Biological Invasions (published)
doi.org
Independence Day: A Poetry to Liberation -- In Memory of USA Independance (Version -1)
Nonvikan Karl-Augustt Alahassa
Leonard Wantchekon
Nathalie Lacelle
Marlène Frigon
Samuel Bassetto
Foutse Khomh
Suljo Linic
Dimitrios Koukoulopoulos
Mylène Bédard
Damien Échevin
Jérôme Théau
Daniel F. Nadeau
David Haziza
Bidossessi R.U. Alahassa
Bakary Manga
J. Tossa
Christiane Rousseau
Maciej Augustyniak
Emmanuel Stip
Julie Carrier
We would like to express few words of Gratitude to USA, as July 04th, 2026, is their Independence day.
2026-07-04
Zenodo (unknown)
doi.org
When Code Authors Are Agents: A Large-Scale Study of Human–Agent Collaboration in Pull Requests
Anthonia Oluchukwu Njoku
Zohreh Sharafi
Foutse Khomh
2026-07-04
ACM Conference on AI-Powered Software (published)
doi.org
AquaGen: Scaling generative models to molecular dynamics precision on thousands of atoms
Emmanuel Bengio
Sanjeev Raja
Yui Tik Pang
Kerstin Klaeser
Cristian Gabellini
Nikhil Shenoy
Francesco Di Giovanni
Prudencio Tossou
We present AquaGen, the first all-atom, explicit solvent, periodic-boundary-condition-aware generative model that produces molecular configu… (see more)rations from the Boltzmann distribution at a fraction of the cost of molecular dynamics (MD). This is in contrast with existing generative models that remove degrees of freedom by operating on coarse-grained, vacuum, or implicit solvent systems. Operating at this resolution allows for post-processing through force field energy evaluations and MD simulations, and enables the prediction of relevant properties in a gray-box manner (as ensemble averages of potential energy evaluations over generated samples). We demonstrate the utility of this paradigm on absolute hydration free energy (AHFE), producing estimates 4-10x faster and with comparable accuracy to standard GPU-based MD. By generating uncorrelated samples from alchemical Boltzmann distributions, we create more accurate, interpretable, and refinable ensemble predictions with calibrated uncertainty estimates, unlike regression methods which are entirely black-box predictors. Our approach also yields predictable benefits from increasing train- and test-time compute, realized by scaling model size and generating more samples, respectively. We believe that this approach demonstrates the utility of high-resolution ensemble generation for free energy estimation, with future potential to replace MD in tasks such as the prediction of lipophilicity, membrane permeability, or absolute binding free energy (ABFE) -- whose grounding and interpretability may be critical for the development of new drugs and materials.
2026-07-02
arXiv (preprint)
doi.org
arxiv.org
Can Model Merging Improve Aggregation in DiLoCo?
Stefan Horoi
Benjamin Therien
Guy Wolf
Eugene Belilovsky
Model merging techniques, which aggregate independently finetuned models into one to combine their capabilities, have become a topic of sign… (see more)ificant interest in recent years, with a broad array of methods having been proposed to tackle this problem. Simultaneously, an emerging trend in distributed learning has been the use of methods such as local SGD and DiLoCo, which greatly reduce communication costs by periodically aggregating the independently trained local models. However, these communication-efficient methods have been shown to degrade in performance relative to the FLOP-matched data-parallel gold standard as the number of independent local models grows and as the number of local training steps before global communication is increased. In this work, we draw an explicit analogy between the pseudo-gradient aggregation step in local SGD/DiLoCo and task arithmetic-based model merging, establishing a straightforward way to utilize merging methods in the context of distributed optimization. We then evaluate multiple state-of-the-art model merging methods in this setting and identify one method in particular, Iso-C, as a promising approach for improving DiLoCo. We find that DiLoCo SGD with Iso-C aggregation outperforms not only simple pseudo-gradient averaging but even the momentum-based DiLoCo, despite lacking a momentum mechanism itself. Building on this finding, we propose IsoLoCo, which adapts Iso-C for distributed training by equipping it with Nesterov momentum. Our empirical evaluations on language model pre-training across varying numbers of local workers show that IsoLoCo significantly outperforms DiLoCo, with the gap between them widening as the number of workers increases. This advantage remains present across model sizes and inner step counts, confirming that merging-inspired aggregation is an effective strategy for low-communication distributed training.
2026-07-02
arXiv (preprint)
doi.org
arxiv.org
When Geometry Aligns: Dihedral Hidden-State Transformations in UNet, ViT, and DiT Architectures
Mojtaba Faramarzi
Alex Lamb
Irina Rish
Diffusion architectures now encompass convolutional UNets as well as transformer-based designs such as Diffusion Transformers (DiTs), inspir… (see more)ed by Vision Transformers (ViTs), yet the effects of structured geometric perturbations within these architectures remain poorly understood. We study this question through a unified framework that applies reflection-based elements of the dihedral group to intermediate hidden states as controlled internal interventions, contrasting geometrically consistent and inconsistent variants. Using activation-level diagnostics, including Self-Consistency Shift (SCS), Activation Mass Scatter (AMS), and Drift, we analyze feature stability and geometric drift. We find that consistent transformations improve stability, while inconsistent ones induce predictable, architecture-specific failures. In the main Stable Diffusion 2.1 U-Net study, we evaluate seven intervention modes over three seeds and complement the internal diagnostics with image-level FID, KID, CLIP score, and LPIPS diversity. Taken together with supporting ViT and controlled DiT analyses, these results establish geometric consistency as a key principle for stable hidden-state interventions in spatially structured vision and diffusion models.
2026-07-02
arXiv (preprint)
doi.org
arxiv.org
LACUNA: A Testbed for Evaluating Localization Precision for LLM Unlearning
Matteo Boglioni
Thibault Rousset
Siva Reddy
Marius Mosbach
Verna Dankers
LLMs memorize sensitive training data, including personally identifiable information (PII), creating a pressing need for reliable post hoc r… (see more)emoval methods. Unlearning has emerged as a promising solution, with state-of-the-art(SOTA) methods often following a localize-first, unlearn-second paradigm that targets specific model parameters. However, existing benchmarks evaluate unlearning solely at the output level, leaving open the question of whether unlearning truly erases knowledge from a model's parameters or merely obfuscates it, a concern reinforced by the success of resurfacing attacks. To bridge this gap, we introduce LACUNA: the first unlearning testbed with ground-truth parameter-level localization. LACUNA injects PII of synthetic individuals into predefined parameters of 1B and 7B OLMo-based models via masked continual pretraining, enabling direct evaluation of whether unlearning targets the weights responsible for knowledge storage. We use LACUNA to benchmark current SOTA unlearning methods and find that, despite strong output-level performance, existing methods are highly imprecise and susceptible to resurfacing attacks. We further show that when localization is successful, even a simple gradient-based unlearning method achieves strong erasure and robustness to resurfacing attacks, highlighting the importance of precise unlearning. We release LACUNA to complement behavioral evaluations and drive further advances in robust, localization-based unlearning.
2026-07-01
arXiv (preprint)
doi.org
arxiv.org
Measure and Integration
Nonvikan Karl-Augustt Alahassa
J. Tossa
Bidossessi R.U. Alahassa
Nathalie Lacelle
Marlène Frigon
Maciej Augustyniak
Dimitrios Koukoulopoulos
Samuel Bassetto
Daniel F. Nadeau
Leonard Wantchekon
Bakary Manga
Victor M. Panaretos
Christiane Rousseau
David Haziza
Julie Carrier
Emmanuel Stip
Mylène Bédard
Bruno Rémillard
Foutse Khomh
Suljo Linic … (see 2 more)
Damien Échevin
Jérôme Théau
We have edited Some lines of Mathematics Notes about Lebesgue Measure and Integration Series for Professor Aboubacar Marcos (abmarcos@yahoo.… (see more)fr), When we were still in Ecole Normale Superieure of Natitingou, Many Professors have contributions, Joel Tossa (joel.tossa@imps-uac.org), (https://archive.org/details/mon-message-a-la-secretaire-d-etat-de-france/joel-tossa%20-%20123x-maxresdefault.pdf, https://archive.org/details/mesure_20260702T182257Z_3_001-aboubacar-marcos/Votre%20Cours%20de%20Mesure%20et%20int%C3%A9gration/).
2026-07-01
Zenodo (unknown)
doi.org
Empowering 2D neural network for 3D medical image segmentation via neighborhood information fusion
Qiankun Li
Xiaolong Huang
Yani Zhang
Bo Fang
Duo Hong
Junxin Chen
2026-06-30
Pattern Recognition (published)
doi.org
Loss Smoothing for Stable Adaptation Under Distribution Shift
Darshan Patil
Ekaterina Lobacheva
Razvan Pascanu
Sarath Chandar
In settings such as fine-tuning and reinforcement learning, neural networks are often adapted under distribution shift. Standard adaptation … (see more)methods typically optimize the target objective directly, inducing an abrupt change from the source training objective. This abrupt transition can distort learned representations, including features that may still be useful for the new task. We investigate whether a more gradual transition can improve adaptation. We propose loss smoothing, a simple approach that interpolates between the source and target training objectives at the start of adaptation. This smooth transition helps to preserve useful features from the source distribution while still enabling the model to specialize to the target distribution. Across controlled supervised shifts, pretrained vision adaptation, offline-to-online and online reinforcement learning, and language model fine-tuning, we find that loss smoothing consistently improves performance, suggesting that smoother objective transitions are a broadly useful tool for model adaptation.
2026-06-30
arXiv (preprint)
doi.org
arxiv.org
The Orders Beyond MLOR (OBM) clause: Enhancing flexibility in retailer–supplier agreements for perishable products
Maria João Santos
Margarida Carvalho
Pedro Amorim
Sara Martins
When shopping for perishable products, consumers typically prefer the freshest items, especially those with a short shelf life. With that in… (see more) mind, retailers establish strict contractual agreements with suppliers to ensure the fulfilment of their orders for perishable products. One key condition in these agreements is the Minimum Life On Receipt (MLOR) rule, which defines the maximum product age that the retailer will accept at full price. In this study, we propose a model that facilitates the negotiation of retailer-supplier terms to increase flexibility. Specifically, we define the share of orders retailers should accept beyond the MLOR at a discounted price. We formulate the problem as a bilevel program considering the individual objectives of the retailer (leader) and the supplier (follower), while also accounting for consumer demand driven by both price and product freshness. To address the bilevel problem, we employ a reformulation-and-decomposition algorithm adapted from the literature. We then compare the supply chain benefits of solving the bilevel program with those of optimising the retailer’s and supplier’s objectives jointly in a centralised approach, as well as to standard contract terms in which products are returned if the supplier does not meet the MLOR requirement. Our results demonstrate that flexible agreements offer significant benefits, with average profit increases of up to 4% for retailers and up to 13% for suppliers. Finally, we provide suggestions for designing new clauses that account for consumer demand variability and retailer’s order frequency.
2026-06-29
International Journal of Production Economics (published)
doi.org