Perception & Reasoning
Fuse visual, spatial, and linguistic input to understand physical environments and identify uncertainty.
In conjunction with ICMI 2026
Human-centric Embodied Multimodal Interaction
An interdisciplinary tutorial on enabling embodied agents to perceive, clarify, personalize, and safely collaborate with humans in the shared physical world.
DayMonday, 5 October 2026
Time09:00 - 12:30
RoomAula magna (UniNa)
Theme
Embodied AI built around people
HEMI bridges vision-language representation, human-agent interaction, and social sciences for trustworthy agents in dynamic real-world environments.
Proactive Interaction
Use verbal and non-verbal cues to clarify ambiguous intent and prevent risks during action.
Personalized Interaction
Adapt agent behavior and communication styles to situated context and individual preferences.
Learning Objectives
View objectives →
What attendees will take away
Embodied perception methods
Understand how MLLMs and VLAs ground multimodal input.
Proactive strategies
Resolve ambiguity through appropriate human-agent dialogue.
MLLM personalization
Tailor behavior, action, and communication to users.
Social signals
Bring gaze, posture, and gestures into embodied AI.
Agenda
Full program →
Tentative half-day schedule
Alessandro VinciarelliMultimodal Social AI
Davide TalonPersonalized human-agent interaction with MLLMs
Yiming WangEmbodied perception and reasoning in human-centric environments
Organizers
Biographies & publications →
Meet the speakers
FBK, Italy
Yiming Wang
Embodied perception, semantic understanding, and navigation/manipulation.
FBK, Italy
Davide Talon
Multimodal learning, vision-language model adaptation, and human-AI interaction.
University of Glasgow, UK
Alessandro Vinciarelli
Computational Social Intelligence and social signal processing.
Supporting Material
Resources for the community
This site is the central hub for supporting material.
Access resources →Reading List
Slides