June 4th / CVPR 2026 / Denver
The First Workshop on
Video Generative Models: Benchmarks and Evaluation
Exploring Challenges and Opportunities in Evaluating and Benchmarking Video Generative Models
Thursday, June 4th, 2026 at 9:00 AM.
Mile High 3B, Colorado Convention Center
Supported by
The rapid advancement of video generative models underscores the critical need for robust evaluation methodologies capable of rigorously assessing instruction adherence, physical plausibility, human fidelity, and creativity. However, prevailing metrics and benchmarks remain constrained, predominantly prioritizing semantic alignment while often overlooking subtle yet critical artifacts, such as structural distortions, unnatural motion dynamics, and weak temporal coherence, that persist even in state-of-the-art systems.
Therefore, the VGBE workshop seeks to pioneer next-generation evaluation methodologies characterized by fine-grained granularity, physical grounding, and alignment with human perception. By establishing multi-dimensional, explainable, and standardized benchmarks, we aim to bridge the gap between generation and assessment, thereby accelerating the maturation of video generative models and facilitating their reliable deployment in real-world applications.
Topics
🏆 Workshop Paper Awards
Recognizing outstanding contributions in workshop paper submissions.
Novel Metrics and Evaluation Methods
- Spatiotemporal & Causal Integrity: Quantifying motion realism, object permanence, and causal logic consistency over time.
- Perceptual Quality Assessment: Learning-based metrics for detecting visual artifacts, hallucinations, and alignment with human subjectivity.
- Explainable Automated Judges: Leveraging Multimodal LLMs (VLMs) for scalable, fine-grained, and interpretable critique.
- Instruction Adherence Metrics: Rigorous evaluation of prompt fidelity, spatial conditioning, and complex constraint satisfaction.
Datasets and Benchmarks
- Narrative & Multi-Shot Suites: Curated datasets assessing character persistence, scene transitions, and long-horizon consistency.
- Physics-Grounded Challenge Sets: Scenarios isolating fluid dynamics, collisions, and kinematic anomalies to stress-test "World Simulators."
- Human Preference Data: Large-scale, fine-grained annotations capturing multi-dimensional judgments (e.g., aesthetics vs. realism).
- Standardized Protocols: Unified data splits and reproducible frameworks to ensure transparent and comparable benchmarking.
Developing video generative applications in vertical domains
- Domain Adaptation & Personalization: Efficient fine-tuning and Low-Rank Adaptation (LoRA) strategies for specialized verticals (e.g., medical, cinematic).
- Simulation for Embodied AI: Leveraging video generative models as world simulators for robotics perception, planning, and Sim2Real transfer.
- Interactive & Human-in-the-Loop: User-centric frameworks incorporating iterative feedback for creative workflows and gaming.
- Immersive 4D Generation: Lifting video diffusion priors to synthesize spatially consistent scenes and dynamic assets for AR/VR environments.
- Deployment Efficiency: Optimizing inference latency, memory footprint, and cost for scalable industrial applications.
Challenges
Submissions will be evaluated on the test set using the metrics defined in the associated paper, with human evaluation conducted for each task as needed.
Image-to-Video Consistent Generation
- Objective: Maintain visual preservation and spatiotemporal consistency from an image and text prompt.
-
Awards:
- 1st Place:$1,000+ Certificate
- 2nd Place:$600+ Certificate
- 3rd Place:$300+ Certificate
- Data Usage: Please follow the Dataset License for data access and usage.
Competition Timeline
| Competition starts | February 19, 2026 |
| Results and Code Submission deadline | April 05, 2026 |
Generic Instructional Video Editing | Website (for more detailed)
- Objective: Edit input videos from natural language instructions while preserving quality and fidelity.
-
Awards:
- Highest Score Award:$500+ Certificate
- Innovation Award:$500+ Certificate
Competition Timeline
| Competition starts | February 20, 2026 |
| Results Submission deadline | April 05, 2026 |
Physics-aware Video Instance Removal | Website (for more detailed)
- Objective: Remove target instances and restore realistic environment dynamics with minimal artifacts.
-
Awards:
- Highest Score Award:$500+ Certificate
- Innovation Award:$500+ Certificate
Competition Timeline
| Competition starts | February 20, 2026 |
| Results Submission deadline | April 05, 2026 |
Keynote Speakers
Organizers
Schedule
| Time | Session | Speaker / Host | Topic / Notes |
|---|---|---|---|
| 9:00 - 9:10 AM | Morning Opening | Organizing Committee | Welcome & Workshop Overview |
| 9:10 - 9:25 AM | Oral Presentation | - | Inferring Dynamic Physical Properties from Video Foundation Models |
| 9:30 - 10:00 AM | Keynote |
Mike Shou
|
Video World Model for Robot Learning |
| 10:00 - 10:30 AM | Keynote |
Yan Wang
|
TBD |
| 10:30 - 10:50 AM | Coffee Break | - | - |
| 10:50 - 11:20 AM | Keynote |
Yaoyao Liu
|
Enable Explicit 3D/4D Controls for Pre-trained Generative Models |
| 11:20 - 11:30 AM | Challenge Winners Ceremony | - | - |
| 11:30 - 11:50 AM | Challenge Winner Solutions | - | TBD |
| 11:50 AM - 1:30 PM | Lunch Break | - | - |
| 1:30 - 1:40 PM | Afternoon Opening | Organizing Committee | - |
| 1:40 - 2:10 PM | Keynote |
Alan Bovik
|
Two Experiments on the Perception of GenAI Pictures |
| 2:10 - 2:40 PM | Keynote |
Zhuang Liu
|
Building and Evaluating Fully Open Generative Models |
| 2:40 - 3:00 PM | Coffee Break | - | - |
| 3:00 - 3:30 PM | Keynote |
Ming-Hsuan Yang
|
Toward World Models: Geometry, View Synthesis, and Visual Reasoning |
| 3:30 - 4:00 PM | Keynote |
Jiajun Wu
|
TBD |
| 4:00 - 4:15 PM | Oral Presentation | - | Physics-Aware Video Instance Removal Benchmark |
| 4:15 - 4:30 PM | Oral Presentation | - | Generative Action Tell-Tales: Assessing Human Motion in Synthesized Videos |
| 4:30 - 4:45 PM | Oral Presentation | - | Risk-Controllable Multi-View Diffusion for Driving Scenario Generation |
| 4:45 - 5:00 PM | Oral Presentation | - | TBD |
| 5:00 - 5:15 PM | Paper Awards Ceremony | Organizing Committee | - |
| 5:15 - 5:30 PM | Closing Remarks & Group Photo | Organizing Committee | - |
Accepted Papers
- T2AV-Compass: Towards Unified Evaluation for Text-to-Audio-Video Generation
- Distilling Geometry Priors for 3D-Consistent Video Generation
- Inferring Dynamic Physical Properties from Video Foundation Models
- Physics-Aware Video Instance Removal Benchmark
- VideoASMR-Bench: Can AI-Generated ASMR Videos Fool VLMs and Humans?
- AIGVE-MACS: Unified Multi-Aspect Commenting and Scoring Model for AI-Generated Video Evaluation
- Objects in Generated Videos Are Slower Than They Appear: Models Suffer Sub-Earth Gravity and Don’t Know Galileo’s Principle…for now
- Tempered Self-Similarity Alignment for Physically Plausible Video Generation
- V-PartSwap: Motion-Consistent Facial Part Transfer in Videos via Alignment-Aware Diffusion
- Generative Action Tell-Tales: Assessing Human Motion in Synthesized Videos
- Risk-Controllable Multi-View Diffusion for Driving Scenario Generation
- Test-Time Domain Adaptation for Interactive Video Generation
- The Evaluation Imperative for Video Generative Models: A Survey on Metrics, Benchmarks, and Trustworthiness