One Day Meeting: BMVA Symposium on World Models
Wednesday 18 November 2026
Chairs: Silpa Vadakkeeveetil Sreelatha (University of Surrey), Adarsh Kappiyath (University of Surrey), Ahmed Bourouis (University of Surrey), Frank Tsu-Fang Lu (University of Surrey), Niladri Shekhar Dutt (University College London)
Queries? contact the Meeting's Organiser Andrew Gilbert here
Invited Speakers
- Amir Bar - Imperial College London, AMI Labs
- Viorica Pătrăucean - Google DeepMind
- Katja Hofmann - Microsoft Research
- Vassia Simaiaki - Wayve
- Niloy Mitra - University College London (UCL) and Adobe Research
- Andrea Vedaldi - University of Oxford
- Amir Ghalamzan - University of Sheffield
Overview
World models are learned predictive models of an environment and its dynamics that can act as an internal simulator for prediction, planning, and policy learning. Progress in world modeling has come from several directions: generative and interactive video models, latent predictive architectures, 3D and 4D scene representations, model-based reinforcement learning, and model predictive control for robots and vehicles. Our speakers span foundations of generative and predictive modelling, video and perception at scale, interactive environments and games, 3D geometry and reconstruction, and robot manipulation and autonomous driving.
Schedule
| Start | End | Title | ||
|---|---|---|---|---|
| 09:10 | 09:20 | Registration & Coffee | ||
| 09:20 | 09:25 | Welcome and opening remarks | ||
| 09:25 | 10:05 | Keynote Talk 1 — Amir Bar | ||
| 10:05 | 10:15 | Lightning Talk 1 - Jonathan Sadeghi — Are the Stochastic Dynamics of Video World Models Physically Calibrated? | ||
| 10:15 | 10:55 | Keynote Talk 2 — Viorica Pătrăucean | ||
| 10:55 | 11:35 | Morning Coffee Break + Morning Poster Session | ||
| 11:35 | 12:15 | Keynote Talk 3 — Katja Hofmann | ||
| 12:15 | 12:25 | Lightning Talk 2 - Farhad Bazyari — Measuring What World Models Forget in Bits | ||
| 12:25 | 13:05 | Keynote Talk 4 — Niloy Mitra | ||
| 13:05 | 13:50 | Lunch | ||
| 13:50 | 14:30 | Keynote Talk 5 — Andrea Vedaldi | ||
| 14:30 | 14:40 | Lightning Talk 3 - Mirlan Karimov — Overcoming the Challenges of Driving World Models in High-Dimensional Semantic Spaces | ||
| 14:40 | 15:20 | Keynote Talk 6 — Amir Ghalamzan | ||
| 15:20 | 16:00 | Afternoon Coffee Break + Afternoon Poster Session | ||
| 16:00 | 16:10 | Lightning Talk 4 - Tom Cashman — 3D Latent Memories for Geometrically-Consistent World Models | ||
| 16:10 | 16:10 | Lightning Talk 5 - Sara Atito Ali Ahmed — Beyond Scale Toward Structured Visual Foundation Models | ||
| 16:20 | 17:00 | Keynote Talk 7 — Vassia Simaiaki | ||
| 17:00 | 17:00 | Closing remarks |
Lightening Talks
8 min each + 2 mins questions
Posters
Please A1 Portrait in size and dimension
Morning Poster Session
- Zican Wang – Neural Voxel Dynamics: Learning Volumetric Feature Advection for 3D Physics in V-JEPA Latent Space
- Aymen Mir – Animatable Humans under Occlusion from YouTube Video
- Xuanya Zhu – How Latent Transition Geometry Matters for Planning in World Models
- Mohamed Hamdi – Handover of Control: Post-Training V-JEPA 2-AC to Finish With MPC What Model-Free RL Cannot
- Moucheng Xu – 4D-VQ-GAN: A World Model for Synthesizing Medical Scans at Any Time Point for Personalized Disease Progression
- Julio Hurtado – Framework for Continual Adaptation Under Physical Distribution Shift
- Ayoub Louaye Bouaziz – HF-BWM: A General Framework for Predictive Modeling of Human Behavior
- Tsun-Yi Yang – World Models for Multi-Agent Systems: Simulation, Compute, and Memory**
Afternoon Poster Session
- Jacob Lin – Forecasting 3D Object Motion with Dense 3D Point Tracks
- Ye Mao – Multi-View World Foundation Model for Robotics Manipulation
- Steven Davenport – Adapt Before Acting: Just-in-Time Policy Updates Through Grounded Imagination
- Andrew Knowles – Qualifying Action-Conditioned JEPA World Models for Quadruped Navigation
- Xuelong An – Gödel-Clifford-Einthoven (G.C.E): Physics Simulator Driven SSL for Electrocardiograms
- Venu Tammabatula – A Patient Is Not a Point: Time-Anchored World Models for Clinical Deterioration in Blood Cancer
- Burak Kizil – From Text to Cinematic Physics: Bridging Action, Geometry, and World Models in Video Generation
- Matthieu Ostertag – World Models for Cross-Machine CNC Transfer under Partial Sensor Overlap
- Kleyton da Costa – Predictive Safety in Action-Conditioned World Models
Important: This is an in-person event, with no virtual attendance option. We kindly ask all presenters to join us at the British Computer Society on the day.
Meeting Location
The meeting will take place at:
British Computer Society (BCS), 25 Copthall Avenue, London EC2R 7BP
Registration
We keep the cost of attending these events as low as possible to ensure no barriers from the whole computer vision community attending. The registration costs are as follows
- All Attendees: £30 Including lunch and refreshments for the day