Xavier Marjou

Antoine Le Borgne, Xavier Marjou, Fanny Parzysz et al.

Extracting information from geographic images and text is crucial for autonomous vehicles to determine in advance the best cell stations to connect to along their future path. Multiple artificial neural network models can address this challenge; however, there is no definitive guidance on the selection of an appropriate model for such use cases. Therefore, we experimented two architectures to solve such a task: a first architecture with chained models where each model in the chain addresses a sub-task of the task; and a second architecture with a single model that addresses the whole task. Our results showed that these two architectures achieved the same level performance with a root mean square error (RMSE) of 0.055 and 0.056; The findings further revealed that when the task can be decomposed into sub-tasks, the chain architecture exhibits a twelve-fold increase in training speed compared to the composite model. Nevertheless, the composite model significantly alleviates the burden of data labeling.

Tangui Le Gléau, Xavier Marjou, Tayeb Lemlouma et al.

In many societal and industrial interactions, participants generally prefer their pure self-interest at the expense of the global welfare. Known as social dilemmas, this category of non-cooperative games offers situations where multiple actors should all cooperate to achieve the best outcome but greed and fear lead to a worst self-interested issue. Recently, the emergence of Deep Reinforcement Learning (RL) has generated revived interest in social dilemmas with the introduction of Sequential Social Dilemma (SSD). Cooperative agents mixing RL policies and Tit-for-tat (TFT) strategies have successfully addressed some non-optimal Nash equilibrium issues. However, this kind of paradigm requires symmetrical and direct cooperation between actors, conditions that are not met when mutual cooperation become asymmetric and is possible only with at least a third actor in a circular way. To tackle this issue, this paper extends SSD with Circular Sequential Social Dilemma (CSSD), a new kind of Markov games that better generalizes the diversity of cooperation between agents. Secondly, to address such circular and asymmetric cooperation, we propose a candidate solution based on RL policies and a graph-based TFT. We conducted some experiments on a simple multi-player grid world which offers adaptable cooperation structures. Our work confirmed that our graph-based approach is beneficial to address circular situations by encouraging self-interested agents to reach mutual cooperation.

Lucas Tamic, Ilan Jaffeux-Cheniout, Xavier Marjou

While large language models (LLMs) show strong performance in natural language understanding and generation, their evaluation and adaptation to domain-specific constraints in telecommunications customer support remain limited. In addition, data sovereignty, regulatory constraints, and the handling of sensitive customer and network information complicate the use of externally hosted foundation models in this domain. We present a systematic study of parameter-efficient fine-tuning (PEFT) using Low-Rank Adaptation (LoRA) applied to Qwen2.5-3B to build a domain-specific conversational assistant. We introduce a combinatorial synthetic data generation approach based on a glossary of 52 industry-specific terms, producing approximately 30,000 training examples across 1,560 distinct problem scenarios via a generative pipeline powered by Gemini 2.0 Flash. We evaluate 16 LoRA configurations by varying hyperparameters and target modules. Our evaluation extends beyond standard metrics by incorporating energy consumption analysis and qualitative assessment using an LLM-as-a-judge framework with GPT-5.2 and Claude 4.5 Sonnet. Results show a clear divergence between quantitative and qualitative performance: models achieving the lowest validation loss do not necessarily obtain the best human-aligned rankings. The best validation loss (0.5024) ranks only 6th-7th in qualitative evaluation, while the worst loss (0.6807) ranks first according to both judges. This work contributes (1) a combinatorial method for synthetic dataset construction, (2) insights into the impact of target module selection for LoRA injection, (3) evidence that validation loss alone is insufficient for selecting fine-tuning configurations in conversational AI, and (4) an energy-performance trade-off analysis for sustainable LLM deployment.

Xavier Marjou

Transcribing spoken audio samples into the International Phonetic Alphabet (IPA) has long been reserved for experts. In this study, we examine the use of an Artificial Neural Network (ANN) model to automatically extract the IPA phonemic pronunciation of a word based on its audio pronunciation, hence its name Generating IPA Pronunciation From Audio (GIPFA). Based on the French Wikimedia dictionary, we trained our model which then correctly predicted 75% of the IPA pronunciations tested. Interestingly, by studying inference errors, the model made it possible to highlight possible errors in the dataset as well as to identify the closest phonemes in French.

Xavier Marjou

To transcribe spoken language to written medium, most alphabets enable an unambiguous sound-to-letter rule. However, some writing systems have distanced themselves from this simple concept and little work exists in Natural Language Processing (NLP) on measuring such distance. In this study, we use an Artificial Neural Network (ANN) model to evaluate the transparency between written words and their pronunciation, hence its name Orthographic Transparency Estimation with an ANN (OTEANN). Based on datasets derived from Wikimedia dictionaries, we trained and tested this model to score the percentage of correct predictions in phoneme-to-grapheme and grapheme-to-phoneme translation tasks. The scores obtained on 17 orthographies were in line with the estimations of other studies. Interestingly, the model also provided insight into typical mistakes made by learners who only consider the phonemic rule in reading and writing.