Sofian Chaybouti

I'm a Research Scientist in the vision team at the Technology Innovation Institute (TII), while finishing my PhD at the Tübingen AI Center, advised by Prof. Hilde Kuehne. I'm also participating in the MIT-IBM Watson Sight and Sound Project.

Prior to this, I finished my Master of Engineering in Applied Mathematics at ENSTA Paris in France and my MVA Master (Mathematics, Vision and Learning) at ENS Paris-Saclay.

Before starting my PhD, I worked for 3 years in industry research teams: at Credit Agricole's Datalab in Montrouge, France, where I mostly focused on NLP and research engines, and at Huawei's Noah's Ark Lab in Paris, working on optimization and Reinforcement Learning.

News

2026 Joined the Technology Innovation Institute (TII) vision team as a Research Scientist.
2026 Falcon Perception and Falcon OCR released — SoTA 0.6B/0.3B early-fusion, autoregressive perception/OCR models.
2026 🏆 SigLino won the Best Paper Award at the A2A Multimodal Workshop @ CVPR 2026, and was selected as a Highlight at CVPR 2026.
2026 SigLino and VisRes Bench accepted at CVPR 2026.
2026 MaskInversion accepted at ICLR 2026.
2025 Joined the Technology Innovation Institute (TII) vision team as a Research Scientist Intern.
2025 LeGrad accepted at ICCV 2025.

Research Papers

My current research focuses on multimodal models, especially Video Understanding, Video-Language modeling, Efficient video representation and efficient Multimodal models.