BEGIN:VCALENDAR VERSION:2.0 PRODID:https://evenements.uqam.ca X-PUBLISHED-TTL:P1W BEGIN:VEVENT UID:34629@https://evenements.uqam.ca DTSTART:20261119T103000Z SEQUENCE:8 TRANSP:OPAQUE URL:https://evenements.uqam.ca/evenements/seminaire-au-dic-perception-token s-for-visual-reasoning/34629?date=2026-11-19_10-30-00 LOCATION:UQAM - Pavillon Président-Kennedy (PK) (201\, avenue du Présiden t-Kennedy\, Montréal ) SUMMARY:Séminaire au DIC: «Perception tokens for visual reasoning» par R anjay Krishna CLASS:PUBLIC DESCRIPTION:Séminaire ayant lieu dans le cadre du Doctorat en informatique cognitive\, en collaboration avec le centre de recherche CRIA               \n\n\n \n\n\nTITRE : Perception tokens for visual reason ing\n\n\n \n\n\nRanjay KRISHNA\n\n\nJeudi le 19 novembre 2026 à 10h30\n\ n\nLocal PK-5115 (Il est possible d'y assister en virtuel en vous inscriv ant ici)                 \n\n\n \n\n\nRÉSUMÉ\n\n\nToda y’s vision-language models can describe images fluently\, yet they still struggle with the spatial and perceptual reasoning required for robotics. I will argue that visual reasoning—reasoning through points\, sketches\ , trajectories\, depth maps\, and other grounded representations—will be come more important than language reasoning for physical agents. It traces progress from benchmarks exposing fundamental perceptual failures to meth ods that teach models to reason directly in visual space. These ideas culm inate in MolmoAct and MolmoAct2\, open vision-language-action models that convert visual plans into robot actions. The resulting systems are interpr etable\, steerable\, deployable zero-shot on affordable hardware\, and com petitive with proprietary models trained on substantially more data—adva ncing the goal of accessible\, general-purpose robots for real-world envir onments.\n\n\n \n\n\nBIOGRAPHIE\n\n\nRanjay KRISHNA is Assistant Professo r in the Paul G. Allen School of Computer Science &\; Engineering at th e University of Washington\, where he co-directs the RAIVN lab. His resear ch bridges computer vision\, natural language processing\, robotics\, and human-computer interaction\, focusing on teaching machines to perceive the visual world and interact with people by drawing on behavioral and social -science frameworks. He previously directed the multimodal and embodied AI team at the Allen Institute for AI.\n\n\n \n\n\nRÉFÉRENCES \n\n\nBigv erdi\, M.\, Luo\, Z.\, Hsieh\, C.-Y.\, Shen\, E.\, Chen\, D.\, Shapiro\, L . G.\, &\; Krishna\, R. (2025). Perception Tokens Enhance Visual Reaso ning in Multimodal Language Models. Conference on Computer Vision and Patt ern Recognition (CVPR).\n\n\nLiu\, B.\, Dong\, Y.\, Wang\, Y.\, Ma\, Z.\, Tang\, Y.\, Tang\, L.\, Rao\, Y.\, Ma\, W.-C.\, &\; Krishna\, R. (2025) . Coarse Correspondences Boost Spatial-Temporal Reasoning in Multimodal L anguage Model. Conference on Computer Vision and Pattern Recognition (CVPR )\, 3783–3792.  \n\nMot-clés : Séminaire\, LLMs\, LATECE UQAM INFORMA TIQUE\, LATECE\, CRIA\, NeuroQAM\, Neurolab\, Département de Neuroscience \, Neurosciences\, Sciences cognitives\, Philosophie\, Sciences cognitive\ , École de langues\, neurosciences cognitives\, Institut des sciences cog nitives\, sciences du langage\, apprentissage machine\, Apprentissage du l angage naturel\, apprentissage profond\, langage automatique\, langage cog nitif\, metacognition\, Cognition humaine\, Cognition\, intelligence artif icielle\, IA \; intelligence artificielle\, société\, IA\, intelligence artificielle\, chatGPT\, enseignement supérieur\, IA\, intelligence artif icielle\, département de linguistique\, Maîtrise en informatique\, maît rise en psychologie\, département de psychologie\, Faculté des sciences de l'UQAM\, Faculté des sciences humaines\, Faculté des sciences\, Dépa rtement d'informatique\, doctorat en informatique cognitive\, doctorat en informatique\n\nPrix : Gratuit\n\n CATEGORIES:Séminaire,Conférence DTSTAMP:20261023T082618Z CREATED:20261013T190958Z LAST-MODIFIED:20261015T110357Z END:VEVENT END:VCALENDAR