NPC to NPC
Social intelligence, perception, communication, collaboration, emotion, and long-term relationships.
ECCV 2026 Workshop
Agent in World: Living Worlds with Interactive Agents
A half-day workshop on interactive intelligence at the intersection of generative world models, embodied agents, visual foundation models, games, simulation, and human-world interaction.
- Venue
- ECCV 2026
- Date
- September 8, 2026, 8:30 - 12:30
Overview
Can an agent live in a world?
Recent breakthroughs in generative world models and generative agents have opened a path toward living worlds: dynamic, persistent environments where agents perceive, communicate, act, and evolve together with the worlds they inhabit.
Agent in World studies the gap between passive generation and static embodiment. Instead of treating environments as visual backdrops or agents as task executors, the workshop frames agents and worlds as reactive, intelligent participants in the same evolving system.
Workshop Scope
We invite researchers from computer vision, graphics, robotics, games, simulation, HCI, machine learning, and AI safety to discuss how visual foundation models can connect pixels, action, interaction, and long-term behavioral coherence.
Interaction Axes
From isolated agents to living systems
NPC to World
Physical intelligence, manipulation, procedural change, and reactive environment dynamics.
Living Systems
Collective populations interacting inside shared worlds with emergent social and physical behaviors.
Massive Interactions
Human intervention, user control, mixed-initiative world editing, and interactive storytelling.
Topics of Interest
- Generative models for NPC behavior, memory, personality, dialogue, and decision-making.
- Generative world models for interactive environments, procedural content, and dynamic scenes.
- Physics-aware environment representations, simulators, datasets, benchmarks, and evaluation protocols.
- Applications in next-generation NPCs, games, social VR, metaverse systems, training, simulation, and AI-driven storytelling.
- Ethics, safety, controllability, and responsible deployment of interactive agents in generated worlds.
Call for Papers
Submission Tracks
Non-proceedings Track
Original short papers, position papers, benchmark proposals, datasets, systems, demos, and early-stage research are welcome. Accepted submissions will be presented at the workshop but will not appear in the official ECCV proceedings.
Proceedings Track
Original research papers seeking formal workshop publication are invited to submit to the proceedings track. Accepted papers in this track will follow the official formatting, review, and camera-ready requirements announced with the submission instructions.
Published Work Track
We welcome recently published or accepted work relevant to interactive agents, world models, games, simulation, embodied AI, and human-world interaction. This track is intended to broaden discussion and connect active communities; submissions will be considered for talks or posters.
Important Dates
- Submission deadline: July 25, 2026 20:59 GMT+9 (July 24, 2026 23:59 AoE)
- Decision notification: August 5, 2026
- Internal camera-ready deadline: August 12, 2026
- Workshop: September 8, 2026, 8:30 - 12:30 at ECCV 2026
Program
Tentative Half-Day Schedule
Opening remarks and workshop vision
Keynote 1 and Keynote 2
Oral session: Contributed papers
Coffee break and Poster session
Keynote 3 and Keynote 4
Industry demo session
Challenge presentation and award
Best paper award and closing remarks
Invited Speakers
Confirmed and tentative keynote speakers
Confirmed
Yilun Du
Assistant Professor, Harvard University
Confirmed
Lianhui Qin
Assistant Professor, UC San Diego
Confirmed
Jinwei Gu
Principal Research Scientist, NVIDIA
Confirmed
Taku Komura
Professor, The University of Hong Kong
Tentative
Haiyang Liu
Alaya Lab (Tokyo)
Organizers
Research across vision, graphics, agents, and worlds
Organizers
Zhixiang Wang
Alaya Lab (Tokyo)
Yichen Peng
Institute of Science Tokyo
Haiyang Liu
Alaya Lab (Tokyo)
Advising Team
Kaipeng Zhang
Alaya Lab (Tokyo)
Erwin Wu
Institute of Science Tokyo
Bo Zheng
Alaya Lab (Tokyo)
Jinwei Gu
NVIDIA
Taku Komura
The University of Hong Kong
Ming-Hsuan Yang
University of California, Merced / Google DeepMind
Contact
Questions?
Please contact the workshop organizers at wangzx1994@gmail.com, haiyangliu1997@gmail.com, or yichen.p.8896@m.isct.ac.jp.