Research map / Video / Video generation
WorldScore: A Unified Evaluation Benchmark for World Generation
Haoyi Duan, Hong-Xing Yu, Sirui Chen, Li Fei-Fei, Jiajun Wu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use 3D scene generation models instead of video models | world generation · WorldScore-Static | WorldScore-Static: CogVideoX-I2V 62.15 -> WonderWorld 72.69 | Improved |
| use best open-source video model (CogVideoX-I2V) instead of closed-source Gen-3 | video generation · WorldScore-Static | WorldScore-Static: 60.71 -> 62.15 | Improved |
| use CogVideoX-I2V instead of Gen-3 | video generation · WorldScore-Dynamic | WorldScore-Dynamic: 57.58 -> 59.12 | Improved |
| extend 3D scene generation to 4D (4D-fy) for dynamics | 3D/4D generation · WorldScore-Static | WorldScore-Static: 72.69 -> 27.98 | Hurt |
| use CogVideoX-T2V (best video camera control) instead of 3D/4D models for camera controllability | video generation · Camera controllability | Camera Ctrl: 40.22 | Hurt |
| use I2V instead of T2V (CogVideoX) for camera controllability | CogVideoX · WorldScore benchmark | Camera Ctrl: 40.22 -> 38.27 | Hurt |
| use I2V instead of T2V (CogVideoX) for static score | CogVideoX · WorldScore-Static | WorldScore-Static: 54.18 -> 62.15 | Improved |
| use I2V instead of T2V (CogVideoX) for motion magnitude | CogVideoX · WorldScore benchmark | Motion Mag: 47.31 -> 26.42 | Hurt |
| larger motion magnitude vs motion smoothness (VideoCrafter1-T2V) | video generation · WorldScore benchmark | Motion Smooth: 75.00 -> 18.87 | Hurt |
| larger motion magnitude vs motion accuracy | video generation · WorldScore benchmark | Motion Acc | No clear effect |
| long sequence (large world) generation instead of small world | video models · WorldScore-Static | WorldScore-Static | Hurt |
| outdoor scenes instead of indoor for video models vs 3D | video models · WorldScore-Static | WorldScore-Static | Hurt |
Closest papers by meaning
- WorldModelBench: Judging Video Generation Models As World Models (NeurIPS 2025 · similarity 0.89)
- WorldCoder-Bench: Benchmarking Physically Grounded 3D World Synthesis (NeurIPS 2026 · similarity 0.89)
- 4DWorldBench: A Comprehensive Evaluation Framework for 3D/4D World Generation Models (CVPR 2026 · similarity 0.88)
- Video-Bench: Human-Aligned Video Generation Benchmark (CVPR 2025 · similarity 0.87)
- LoopNav: Benchmarking Spatial Consistency in World Models (NeurIPS 2026 · similarity 0.87)
- WonderWorld: Interactive 3D Scene Generation from a Single Image (CVPR 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
