CV
PhD student in the ComLearn team (GIPSA-lab & Inria Grenoble), working on multimodal and social grounding of language models to study language acquisition.
Contact Information
| Name | Théo Charlot |
| Professional Title | PhD Student |
| theocyrilcharlot@gmail.com | |
| Location | Grenoble, France |
Professional Summary
Research interests in human-inspired machine learning, developmentally-plausible machine learning, and human reasoning.
Experience
-
2026 - present Grenoble, France
PhD Student, ComLearn team
GIPSA-lab & Inria Grenoble
How do children learn language by looking around and listening to their noisy everyday life? Multimodal and social grounding of language models, supervised by Thomas Hueber, Stéphane Lathuilière and Laurent Girin.
-
2025 - 2026 Paris, France
Research Engineer, CoML & LAAC teams
École Normale Supérieure (ENS Ulm)
Who speaks to whom? Finetuning BabyHuBERT for child-directed speech (CDS) classification of adult speech.
-
2025 - 2025 Paris, France
M2 Research Intern, CoML team
École Normale Supérieure (ENS Ulm)
Who speaks when? Pretraining and fine-tuning BabyHuBERT, a speech representation model trained on 13k hours of multilingual child-centered egocentric audio for segmenting speakers.
-
2024 - 2024 Paris, France
M1 Research Intern, ALMAnaCH team
Inria Paris
Evaluation of LLMs’ ability to accurately retrieve precise information in long dialogs. Experiments with inference approaches to tackle context limits (RAG, ChainOfThought, InfiniTransformers, MEMGpt).
-
2023 - 2023 Software Engineer
TNS Mars
In charge of implementing a lightweight AI-assisted video annotation tool using OpenCV.
Education
-
2023 - 2025 Master Degree in Computer Science
Nantes Université
Apprentissage et Traitement Automatique de la Langue (ATAL)
- M2: Language Modeling, Advanced NLP, Deep Learning, Scientific Communication
- M1: NLP, LLMs, LLM fine-tuning, Unsupervised Learning, Metaheuristics, Data Science, Algorithmics, Big Data
- Highest Honors (rank 1st/14)
-
2018 - 2022 Paris, France
Bachelor
Sorbonne Université
Computer Science
- Reinforcement Learning, Supervised Learning, Data Science, Algorithmics, Web Fullstack, Linux Systems
Awards
-
2024 Travel Grant
ATALA association
Awarded for participation in DEFT2024 @ TALN2024, ExBERT.
Publications
-
2026 Context-aware child-directed speech detection from long-form recordings
Interspeech
-
2026 BabyHuBERT: Multilingual Self-Supervised Learning for Segmenting Speakers in Child-Centered Long-Form Recordings
Interspeech
-
2026 Frame of Reference: Addressing the Challenges of Common Ground Representation in Situational Dialogs
Findings of ACL
Projects
-
Summarization of U.S. Supreme Court Opinions
Frugal on-device summarization for very long documents (100k+ tokens) with quantized models (Llama 1B Q4_K_M) and creation of an audio dataset (2k+ hours) for audio-only argument opinions.
-
CEM-ERL
Development and analysis of a deep RL algorithm combining a SOTA population and gradient based algorithm (CEM-RL) with evolutionary methods (ERL).