We are happy to share that two of our papers have been accepted at NeurIPS 2026.Learning to Evolve Scenes: Reasoning about Human Activities with Scene Graphs (GLEN) — here we introduce Activity-Conditioned Graph Edit Forecasting: given a scene graph and a textual description of an
activity, GLEN predicts the sequence of node and edge edits that transform the scene as the activity unfolds, turning
scene graphs into a common language for long-horizon reasoning, scene evolution and compositional video retrieval.A General Concept-based Decomposition for Vision–Language Embeddings — a collaboration with Pascal Frossard's
group at EPFL, led by Simone, proposing a general way to decompose vision–language embeddings into interpretable
concepts.Congratulations to all the authors!
