This paper introduces a new training strategy to improve speech dereverberation systems in an unsupervised manner using only reverberant speech. Most existing algorithms rely on paired dry/reverberant data, which is difficult to obtain. Our approach uses limited acoustic information, like the reverberation time (RT60), to train a dereverberation system. Experimental results demonstrate that our method achieves more consistent performance across various objective metrics than the state-of-the-art.

This paper presents the work carried out by the O_FT team, joint with Orange and Ouest-France, on adapting language models to the defense domain as part of the EvalLLM2025 challenge. This work focused on adapting the \texttt{Mistral-7B-Instruct-v0.3} model using classical techniques of continued pre-training and instruction-tuning. The core of our efforts is based on collecting, generating, and selecting data for these two stages as well as for model evaluation. Experiments show that our adapted models have better domain-specific knowledge and improved domain-specific task processing skills, along with comparable (or even superior) performance on general knowledge and skills. Considering the carbon footprint of our adaptations, this work demonstrates the feasibility of domain adaptation for relatively small models. -- Ce document présente les travaux réalisés par l'équipe O_FT conjointe à Orange et Ouest-France sur l'adaptation de modèles de langue au domaine de la défense dans le cadre du challenge EvalLLM2025. Ces travaux se sont concentrés sur l'adaptation du modèle \texttt{Mistral-7B-Instruct-v0.3} avec des techniques classiques de poursuite du pré-entraînement et d'affinage sur instructions. L'essentiel de nos travaux a porté sur la constitution, génération et sélection de données pour ces deux étapes ainsi que pour l'évaluation des modèles. Les expériences montrent que nos modèles adaptés ont de meilleures de connaissances de fond et une meilleure capacité de traitement de tâches sur le domaine de la défense, ainsi que des performances comparables (voire supérieures) sur des connaissances ou capacités généralistes. Mis au regard des empreintes carbones de nos adaptations, ces travaux démontrent ainsi la viabilité de l'adaptation à un domaine de modèles relativement petits.

Last week, to formally consecrate the beginning of the holiday season, my fiancée threw on Christmas With the Kranks. The wretched 2004 Tim Allen film notched an impressively bad 5 percent on Rotten Tomatoes, with most critics complaining about the screenplay's awkward marriage of lifeless Middle American sentimentality with the sort of visual gags you might find in Progressive commercials. Christmas With the Kranks is not making the Criterion Collection anytime soon, and I think it's fair to say that there are better ways to spend a winter evening, but if you value the season like we do--and intend to have a Christmas movie on-screen at all times until New Year's--then you must ration the heavy hitters of the genre for the premium slots on the calendar. Or, in other words, if you want to watch Home Alone on Dec. 22, then you might be forced to spend a night with Tim Allen on Dec. 5. We all know what the classics are. The Mount Rushmore of Christmas movies, at least according to mainstream millennial opinion, are Elf, A Charlie Brown Christmas, the Chuck Jones–animated How the Grinch Stole Christmas, and the aforementioned Home Alone.

Motivated by the cost heterogeneity in experimentation across different alternatives, we study the Best Arm Identification with Resource Constraints (BAIwRC) problem. The agent aims to identify the best arm under resource constraints, where resources are consumed for each arm pull. We make two novel contributions. We design and analyze the Successive Halving with Resource Rationing algorithm (SH-RR). The SH-RR achieves a near-optimal non-asymptotic rate of convergence in terms of the probability of successively identifying an optimal arm. Interestingly, we identify a difference in convergence rates between the cases of deterministic and stochastic resource consumption.

In this work, we proposed a new dynamic distributed planning approach that is able to take into account the changes that the agent introduces on his set of actions to be planned in order to take into account the changes that occur in his environment. Our approach fits into the context of distributed planning for distributed plans where each agent can produce its own plans. According to our approach the generation of the plans is based on the satisfaction of the constraints by the use of the genetic algorithms. Our approach is to generate, a new plan by each agent, whenever there is a change in its set of actions to plan. This in order to take into account the new actions introduced in its new plan. In this new plan, the agent takes, each time, as a new action set to plan all the old un-executed actions of the old plan and the new actions engendered by the changes and as a new initial state; the state in which the set of actions of the agent undergoes a change. In our work, we used a concrete case to illustrate and demonstrate the utility of our approach.

Expectation-Maximization (EM) algorithm is a widely used iterative algorithm for computing maximum likelihood estimate when dealing with Gaussian Mixture Model (GMM). When the sample size is smaller than the data dimension, this could lead to a singular or poorly conditioned covariance matrix and, thus, to performance reduction. This paper presents a regularized version of the EM algorithm that efficiently uses prior knowledge to cope with a small sample size. This method aims to maximize a penalized GMM likelihood where regularized estimation may ensure positive definiteness of covariance matrix updates by shrinking the estimators towards some structured target covariance matrices. Finally, experiments on real data highlight the good performance of the proposed algorithm for clustering purposes.

Generating user activity is a key capability for both evaluating security monitoring tools as well as improving the credibility of attacker analysis platforms (e.g., honeynets). In this paper, to generate this activity, we instrument each machine by means of an external agent. This agent combines both deterministic and deep learning based methods to adapt to different environment (e.g., multiple OS, software versions, etc.), while maintaining high performances. We also propose conditional text generation models to facilitate the creation of conversations and documents to accelerate the definition of coherent, system-wide, life scenarios.

As an implementation of the Nystr\"{o}m method, Nystr\"{o}m computational regularization (NCR) imposed on kernel classification and kernel ridge regression has proven capable of achieving optimal bounds in the large-scale statistical learning setting, while enjoying much better time complexity. In this study, we propose a Nystr\"{o}m subspace learning (NSL) framework to reveal that all you need for employing the Nystr\"{o}m method, including NCR, upon any kernel SVM is to use the efficient off-the-shelf linear SVM solvers as a black box. Based on our analysis, the bounds developed for the Nystr\"{o}m method are linked to NSL, and the analytical difference between two distinct implementations of the Nystr\"{o}m method is clearly presented. Besides, NSL also leads to sharper theoretical results for the clustered Nystr\"{o}m method. Finally, both regression and classification tasks are performed to compare two implementations of the Nystr\"{o}m method.

Clinical decision support systems combine knowledge and data from a variety of sources, represented by quantitative models based on stochastic methods, or qualitative based rather on expert heuristics and deductive reasoning. At the same time, case-based reasoning (CBR) memorizes and returns the experience of solving similar problems. The cooperation of heterogeneous clinical knowledge bases (knowledge objects, semantic distances, evaluation functions, logical rules, databases...) is based on medical ontologies. A multi-agent decision support system (MADSS) enables the integration and cooperation of agents specialized in different fields of knowledge (semiology, pharmacology, clinical cases, etc.). Each specialist agent operates a knowledge base defining the conduct to be maintained in conformity with the state of the art associated with an ontological basis that expresses the semantic relationships between the terms of the domain in question. Our approach is based on the specialization of agents adapted to the knowledge models used during the clinical steps and ontologies. This modular approach is suitable for the realization of MADSS in many areas.

Under the project Maccoy Critical, we would like to train individuals, in virtual environments, to handle critical situations such as dilemmas. These latter refer to situations where there is no ``good'' solution. In other words, situations that lead to negative consequences whichever choice is made. Our objective is to use Knowledge Models to extract necessary properties for dilemmas to emerge. To do so, our approach consists in developing a Scenario Orchestration System that generates dilemma situations dynamically without having to write them beforehand. In this paper we present this approach and expose a proof of concept of the generation process.