Le CRIL en bref
Le Centre de Recherche en Informatique de Lens (CRIL UMR 8188) est un laboratoire de l’Université d’Artois et du CNRS dont la thématique de recherche fédératrice concerne l'intelligence artificielle et ses applications. Il regroupe près de 70 membres : chercheurs, enseignants-chercheurs, doctorants et personnels administratifs et techniques.
Le CRIL participe à la Confédération Européenne de Laboratoires en Intelligence Artificielle CAIRNE et à l'alliance régionale humAIn. Il bénéficie du soutien du Ministère de l’Enseignement Supérieur et de la Recherche, du CNRS, de l’Université d’Artois et de la région Hauts de France.
Le CRIL est localisé sur deux sites à Lens : la faculté des sciences Jean Perrin et l’IUT.
Actualités (RSS)
Séminaire Séminaire Christophe Cerisara-LORIA
LLM cost reduction: a loss landscape point of view
5 févr. 2026 - 14:00In this talk, I will first briefly introduce our recent contributions to LLM training in Nancy, and then I will summarize a selection of the research works that have been carried out in our team to reduce the cost of LLM training, in particular with LLM compression, incremental training, PEFT and weakly-supervised training. All these approaches can be interpreted through a common lens: the loss landscape. I will adopt this point of view as a common thread to tie several of our cost reduction approaches together in my presentation.
Séminaire Séminaire de Michael A. Müller
On argumentation and dialogue
8 janv. 2026 - 14:00In practice, argumentation is often dialogical. We argue mostly not in isolation but with others by exchanging utterances. In contrast, computational approaches that look at argumentation in terms of argumentation frameworks (abstract, bipolar, structured, …) consider argumentation as a structure to be evaluated. In this talk, I first want to to briefly show how the dialogical character of argumentation shows itself in core parts of informal argumentation studies. Then I will discuss how the two perspectives on argumentation can supplement each other, by focusing on two ways and situations where argumentation and dialogues interact: when evaluating inference-claims and in advisory dialogues.
Séminaire Séminaire de Nicolas Schwind
Iterated Belief Change as Learning
4 déc. 2025 - 14:00In this work, we show how the class of improve- ment operators – a general class of iterated belief change operators – can be used to define a learning model. Focusing on binary classification, we present learning and inference algorithms suited to this learning model and we evaluate them empirically. Our findings highlight two key insights: first, that iterated belief change can be viewed as an effective form of online learning, and second, that the well-established axiomatic foundations of belief change operators offer a promising avenue for the axiomatic study of classification tasks.
Cinq papiers acceptés à AAAI'26
Les papiers suivants seront présentés à la 40ème conférence AAAI annuelle d’intelligence artificielle (AAAI'26) : Aperiodic Tiling and Rhythmic Canons: A CP Journey Guillaume Derval, Christophe Lecoutre* Generalizing Analogical Inference from Boolean to Continuous Domains Francisco Cunha, Yves Lepage, Miguel Couceiro, Zied Bouraoui. Targeting in Multi-Criteria Decision Making Nicolas Schwind, Patricia Everaere, Sébastien Konieczny, Emmanuel Lonca Truth-Tracking Evaluation in Opinion-Based Argumentation Juliete Rossie, Jérôme Delobelle, Sébastien Konieczny, Srdjan Vesic Structure-Aware Encodings of Argumentation Properties for Clique-width