Mustapha Bounoua

I am a research scientist interested in how models learn and generate structure from multimodal data. My research lies at the intersection of generative modelling, multimodal learning, and information theory, with applications to language, vision, and scientific data. I am particularly interested in multimodal alignment and in principled ways of understanding and improving modern foundation models.

news

Jan 20, 2026	Our paper TENDE: Transfer Entropy Neural Diffusion Estimation has been accepted at AISTATS 2026 !
Jan 20, 2026	My thesis manuscript is available online Harnessing Multimodality : Diffusion based Generative Modeling and Information Estimation
May 1, 2025	Our paper Learning to Match Unpaired Data with Minimum Entropy Coupling has been accepted at ICML 2025 !

Selected Publications

ArXiv

Improved Sampling Schedules for Discrete Diffusion Models

Alberto Foresti^*, Mustapha Bounoua^*, Giulio Franseze, and 2 more authors

In Preprint, 2026

Preprint

PDF

AISTATS

TENDE: Transfer Entropy Neural Diffusion Estimation

Simon Pedro Galeano Munoz, Mustapha Bounoua, Giulio Franzese, and 2 more authors

In AISTATS, 2026

Poster

Bib PDF

@inproceedings{munoz,
  title = {TENDE: Transfer Entropy Neural Diffusion Estimation},
  author = {Munoz, Simon Pedro Galeano and Bounoua, Mustapha and Franzese, Giulio and Michiardi, Pietro and Filippone, Maurizio},
  booktitle = {AISTATS},
  presentation_type = {Poster},
  year = {2026},
  address = {Tangier, Morocco},
}

Sorbonne

Harnessing Multimodality : Diffusion based Generative Modeling and Information Estimation

Mustapha Bounoua

Sorbonne Université, Jul 2025

Phd thesis

Bib PDF

@phdthesis{bounoua:tel-05416294,
  title = {{Harnessing Multimodality : Diffusion based Generative Modeling and Information Estimation}},
  author = {Bounoua, Mustapha},
  school = {{Sorbonne Universit{\'e}}},
  year = {2025},
  presentation_type = {PhD Thesis},
  month = jul,
}

ICML

Learning to Match Unpaired Data with Minimum Entropy Coupling

Mustapha Bounoua, Giulio Franzese, and Pietro Michiardi

In International Conference on Machine Learning, Jul 2025

Poster

Bib PDF Code Slides

@inproceedings{bounoua2025learningmatchunpaireddata,
  title = {Learning to Match Unpaired Data with Minimum Entropy Coupling},
  author = {Bounoua, Mustapha and Franzese, Giulio and Michiardi, Pietro},
  booktitle = {International Conference on Machine Learning},
  year = {2025},
  presentation_type = {Poster},
}

ICML

SΩI: Score-based O-INFORMATION Estimation

Mustapha Bounoua, Giulio Franzese, and Pietro Michiardi

In International Conference on Machine Learning, Jul 2024

Oral top 1.5%

Bib PDF Code Slides

@inproceedings{bounoua2024somegai,
  title = {SΩI: Score-based O-INFORMATION Estimation},
  author = {Bounoua, Mustapha and Franzese, Giulio and Michiardi, Pietro},
  year = {2024},
  booktitle = {International Conference on Machine Learning},
  presentation_type = {Oral Top 1.5\%},
}

Entropy

Multi-Modal Latent Diffusion

Mustapha Bounoua, Giulio Franzese, and Pietro Michiardi

In Entropy, Deep Generative Modeling: Theory and Applications, Jul 2024

Bib PDF Code

@inproceedings{bounoua2023multimodal,
  author = {Bounoua, Mustapha and Franzese, Giulio and Michiardi, Pietro},
  title = {Multi-Modal Latent Diffusion},
  journal = {Entropy},
  booktitle = {Entropy, Deep Generative Modeling: Theory and Applications},
  volume = {26},
  number = {4},
  year = {2024},
}

ICLR

MINDE: Mutual Information Neural Diffusion Estimation

Giulio Franzese, Mustapha Bounoua, and Pietro Michiardi

In The Twelfth International Conference on Learning Representations, Jul 2024

Poster

Bib PDF Code

@inproceedings{franzese2023minde,
  title = {MINDE: Mutual Information Neural Diffusion Estimation},
  author = {Franzese, Giulio and Bounoua, Mustapha and Michiardi, Pietro},
  booktitle = {The Twelfth International Conference on Learning Representations},
  year = {2024},
  presentation_type = {Poster},
}

Neurips

Masked multi-time diffusion for multi-modal generative modeling

Mustapha Bounoua, Giulio Franzese, and Pietro Michiardi

In NeurIPS 2023, Diffusion Models Workshop, Jul 2023

Workshop

Bib PDF

@inproceedings{EURECOM+7540,
  author = {Bounoua, Mustapha and Franzese, Giulio and Michiardi, Pietro},
  title = {Masked multi-time diffusion for multi-modal generative modeling},
  booktitle = {NeurIPS 2023, Diffusion Models Workshop},
  year = {2023},
  address = {New Orleans},
  presentation_type = {Workshop}
}

* means equal contribution