Maxime Dassen

Hi, I'm Maxime (she/her). I am a second-year PhD student at the IRLab of the University of Amsterdam, advised by Andrew Yates, where I work on foundation models and how they acquire, represent, and use information across modalities. Specifically, my focus is on how these models can be made more interpretable, controllable, and reliable.

I was a visiting student at the Johns Hopkins University HLTCOE as part of the SCALE 2025 program, where I worked on hallucination detection of LLM citations, and the SCALE 2026 program, where I worked on multimodal grounding.

Ultimately, I believe that understanding model internals and how their decisions and outputs form is essential for building AI systems that people can genuinely trust.

Last updated at October 5th 2026

profile photo

Selected Research

I'm interested in machine learning, deep learning, generative AI, and multimodal learning.

Preprint
Grounding on Request: Learning Modality Control in Vision-Language Models
Maxime Dassen, Alexander Martin, Andrew Yates, Reno Kriz, William Walden
arXiv preprint, 2026
arXiv code

We introduce different methods to control which modality a vision-language model grounds its answer in when an image and its text disagree.

ECIR '2026
FACTUM: Mechanistic Detection of Citation Hallucination in Long-Form RAG
Maxime Dassen, Rebecca Kotula, Kenton Murray, Andrew Yates,
Dawn Lawrie, Efsun Kayi, James Mayfield, Kevin Duh
European Conference on Information Retrieval (ECIR), 2026
arXiv code

We introduce FACTUM, a mechanistic framework that detects citation hallucinations in long-form RAG by identifying scale-dependent signatures in transformer pathways, outperforming state-of-the-art baselines by up to 37.5%.


Template adjusted from this website.