PhD student · Inria Sophia-Antipolis

Nabyl Quignon

I study how image and video generative models work, and how to control them.

About

I'm a PhD student at Inria, curious about how generative models work.

Trained on vast amounts of data, these models learn internal representations of concepts, spatial structure and even geometry. My goal is to understand these representations well enough to control models directly, for example deciding which region of an image receives a given concept, without retraining or fine-tuning the model. This is cheaper, avoids the loss of capabilities that fine-tuning can cause, and allows more precise control than conditioning alone.

I'm also interested in representation learning, and in how to help diffusion models learn better representations that lead to better generation.

During my free time, I mainly play guitar on my pink Strat, listen to a lot of music of all genres, discover new places and read novels.

Publications

  1. NeurIPS 2026

    Few-Shot Visual Concept Extraction for Steering Diffusion Transformers

    Nabyl Quignon, Antitza Dantcheva

    Project
  2. CVPR 2026 · Findings

    THEval: Evaluation Framework for Talking Head Video Generation

    Nabyl Quignon, Baptiste Chopin, Yaohui Wang, Antitza Dantcheva

    Project Paper arXiv
  3. WACV 2027

    MoSTE: Mixture of Spatio-Temporal Experts for Deepfake Detection in Videos

    Charbel Yahchouchi*, Nabyl Quignon*, Noemi Roggero, Laurent Saroul, Antitza Dantcheva

    * Equal contribution

    Coming soon
  4. ECCV 2026 · ABAW WorkshopOral

    BEACON: Behavior and Appearance Control for Subject-Specific Video Generation

    Baptiste Pokrzywa, Nabyl Quignon, Yara Bahram, Muhammad Osama Zeeshan, Antitza Dantcheva, Eric Granger

    arXiv
  5. IEEE T-BIOM 2026

    Beyond Real versus Fake: Towards Intent-Aware Video Analysis

    Saurabh Atreya, Nabyl Quignon, Baptiste Chopin, Abhijit Das, Antitza Dantcheva

    arXiv

Experience

  1. Mar 2025 → present

    PhD in Computer Science

    Inria · Sophia-Antipolis, France

    Research on the internal representations of image and video generative models, and on using them to control generation.

  2. Jan → Jul 2024

    Research Intern

    Dassault Systèmes · Paris, France

    Multimodal representation learning for BRep CAD models, linking 3D geometry with text and images. This work led to two pending patent applications.

  3. Apr → Jul 2022

    Software Engineer Intern

    Socomate International · Paris, France

    Interpreted, optimized and implemented in C++ the mathematical formulas used to compute focal laws for a multi-element ultrasonic transducer.

Teaching

  1. 2025–26 · 2026–27

    Teaching Assistant · Applied AI

    Polytech Nice Sophia · Sophia-Antipolis, France

    Lab sessions on generative adversarial networks and diffusion models, for master's students.

Contact