Scientific meeting 2025

On the 21th of January 2025, SAIF organized a scientific day. Each project member had the opportunity to present their work, have productive discussions on joint collaborations (and share a good meal!).

Participants

ANTIQUE, Inria Paris

• Alessandro De Palma
• Caterina Urban

LaBRI, Université de Bordeaux    

• Nathanaël Fijalkow (online)
• Roman Kniazev

LIX, Institut Polytechnique de Paris

• Constantin Enea
• Eric Goubault
• Jérémy Dubut (afternoon only)
• Sergio Mover
• Sylvie Putot

LMF, Université Paris-Saclay

• Serge Haddad (online)
• Benedikt Bollig
• Kerian Thuillier
• Lina Ye

LSL, CEA-List

• Zakaria Chihani
• Sasha Cuau
• Dorin Doncenco
• Julien Girard-Satabin
• Augustin Lemelse
• Romain Xu-Darme    

Inria Rennes

• Eric Fabre

TAU, Inria Saclay

• Guillaume Charpiat

Others

• Blaise Genest (online)
• Liao Yuke (online)

Schedule

        9h30 – 10h00: Project Status and Coordination
        10h00 – 10h30: Collecte ANR 
        10h30 – 11h00: AI Action Summit & PEPR IA Days 
        11h00 – 11h25: Talk by Blaise Genest
        “More Accurate NN Verification than 𝜶β-CROWN: Root Cause for Imprecisions and Neuron Importance Ranking”
        11h25 – 11h50: Talk by Eric Goubault & Sylvie Putot
        “A Zonotopic Dempster-Shafer Approach to the Quantitative Verification of Neural Networks”
        11h50 – 12h00: Talk by Sasha Cuau, 
        “Formal Verification of Transformers”

        12h00 – 14h30: Lunch (restaurant “Le Perraudin”, 157 Rue Saint-Jacques, 75005 Paris)

        14h30 – 14h55: Talk by Roman Kniazev
        “Programmatic Reinforcement Learning: Why and How?”
        14h55 – 15h20: Talk by Eric Fabre
        “An Experimental Journey into Soft Automata”
        15h20 – 15h30: Talk by Dorin Doncenco, 
        “Formal and Actionable Explainable AI”
        15h30 – 15h55: Talk by Julien Girard-Satabin, 
        “Specification and Verification of Safety Properties of AI Components”
        15h55 – 16h20: Talk by Sergio Mover
        “Reachability-aware Abstractions in Hierarchical Reinforcement Learning”
        16h20 – 16h45: Talk by Alessandro De Palma
        “Neural Network Verification: Beyond Proving Norm-bounded Adversarial