Publications

You can also find my articles on my Google Scholar profile.

Conference Papers

Key Ingredients for EEG Foundation Models: What Works - and What Doesn’t

Published in EUSIPCO, 2026

Giulia Lioi, Yassine El Ouahidi, Jonathan Lys, Philipp Thölke, Bastien Pasdeloup, Vincent Gripon, Nicolas Farrugia, Karim Jerbi

This paper analyzes the core methodologies and architectural choices behind EEG foundation models to identify what works and what doesn’t.

Recommended citation: Lioi, et al. (2026). "Key Ingredients for EEG Foundation Models: What Works - and What Doesn't." In 2026 34th European Signal Processing Conference (EUSIPCO). IEEE.

How Transferable Are EEG Foundation Models? A Case Study on Sleep Staging

Published in EUSIPCO, 2026

Alix Lamouroux, Lucas Benoit, Jonathan Lys, Nicolas Farrugia, Giulia Lioi

This paper evaluates the transferability of EEG foundation models using sleep staging as a case study.

Recommended citation: Lamouroux, et al. (2026). "How Transferable Are EEG Foundation Models? A Case Study on Sleep Staging." In 2026 34th European Signal Processing Conference (EUSIPCO). IEEE.

Residual Connections and the Causal Shift: Uncovering a Structural Misalignment in Transformers

Published in EUSIPCO, 2026

Jonathan Lys, Vincent Gripon, Bastien Pasdeloup, Axel Marmoret, Lukas Mauch, Fabien Cardinaux, Ghouthi Boukli Hacene

This paper identifies a residual-path input-output alignment shift in autoregressive Transformers and proposes residual attenuation to mitigate it.

Recommended citation: Lys, et al. (2026). "Residual Connections and the Causal Shift: Uncovering a Structural Misalignment in Transformers." In 2026 34th European Signal Processing Conference (EUSIPCO). IEEE.
Download Paper | Download Bibtex

Inner Loop Inference for Pretrained Transformers: Unlocking Latent Capabilities Without Training

Published in EUSIPCO, 2026

Jonathan Lys, Vincent Gripon, Bastien Pasdeloup, Axel Marmoret, Lukas Mauch, Fabien Cardinaux, Ghouthi Boukli Hacene

This paper proposes inference-time inner looping to extend computation in frozen pretrained Transformers without training.

Recommended citation: Lys, et al. (2026). "Inner Loop Inference for Pretrained Transformers: Unlocking Latent Capabilities Without Training." In 2026 34th European Signal Processing Conference (EUSIPCO). IEEE.
Download Paper | Download Bibtex

TENSLORA: Tensor Alternatives for Low-Rank Adaptation

Published in ICASSP, 2026

Axel Marmoret, Reda Bensaid, Jonathan Lys, Vincent Gripon, François Leduc-Primeau

This paper introduces TensLoRA, a unified framework for tensor-based low-rank adaptations in Transformers.

Recommended citation: Marmoret, et al. (2025). "TensLoRA: Tensor Alternatives for Low-Rank Adaptation." In 2026 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE.
Download Paper | Download Bibtex

REVE: A Foundation Model for EEG – Adapting to Any Setup with Large-Scale Pretraining on 25,000 Subjects

Published in NeurIPS, 2025

Yassine El Ouahidi, Jonathan Lys, Philipp Thölke, Nicolas Farrugia, Bastien Pasdeloup, Vincent Gripon, Karim Jerbi, Giulia Lioi

This paper presents REVE, a foundation model for EEG that generalizes across diverse setups using large-scale pretraining.

Recommended citation: El Ouahidi, et al. (2025). "REVE: A Foundation Model for EEG Adapting to Any Setup with Large-Scale Pretraining on 25,000 Subjects." In Advances in Neural Information Processing Systems (NeurIPS 2025).
Download Paper | Download Slides | Download Bibtex

FICUS: Few-shot Image Classification with Unsupervised Segmentation

Published in EUSIPCO, 2024

Jonathan Lys, Frédéric Lin, Clément Béliveau, Jules Decaestecker, Yassir Bendou, Aymane Abdali, Bastien Pasdeloup

This paper tackles ambiguous few-shot image classification problems with unsupervised segmentation.

Recommended citation: Lys, et al. (2024). "FICUS: Few-shot Image Classification with Unsupervised Segmentation." In 2024 32nd European Signal Processing Conference (EUSIPCO)(pp. 1791-1795). IEEE.
Download Paper | Download Slides | Download Bibtex

Preprints

D5P4: Partition Determinantal Point Process for Diversity in Parallel Discrete Diffusion Decoding

Published in arXiv, 2026

Jonathan Lys, Vincent Gripon, Bastien Pasdeloup, Axel Marmoret, Lukas Mauch, Fabien Cardinaux, Ghouthi Boukli Hacene

This paper introduces D5P4, a beam-style decoding method for parallel discrete diffusion that balances quality and diversity.

Recommended citation: Lys, et al. (2026). "D5P4: Partition Determinantal Point Process for Diversity in Parallel Discrete Diffusion Decoding." arXiv preprint arXiv:2603.19146.
Download Paper | Download Bibtex