Research map / Video / Video generation
VBench++: Comprehensive and Versatile Benchmark Suite for Video Generative Models
Ziqi Huang, Fan Zhang, Xiaojie Xu, Yinan He, Jiashuo Yu, Ziyue Dong, Qianli Ma, Nattapol Chanpaisit, Chenyang Si, Yuming Jiang, Yaohui Wang, Xinyuan Chen and 5 more
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 |
|---|---|---|---|
| favor temporal consistency over motion (Gen-2 vs VideoCrafter-0.9) | T2V models · VBench T2V | Dynamic Degree: 18.89% vs 89.72% | Mixed |
| static scenes inflate consistency (LaVie) | LaVie · VBench T2V | Dynamic Degree: 49.72% | Hurt |
| use T2I models instead of T2V models for compositionality | T2V models · VBench T2V | Multiple Objects / Spatial Relationship | Improved |
| add more Human data (26% of WebVid-10M) for the Human category | T2V models · VBench T2V | Human category performance: 26% | No clear effect |
| rely on Food data (11% of WebVid-10M) for aesthetics | WebVid-10M · VBench T2V | Aesthetic Quality: 11% | Improved |
| align I2V outputs more closely with input image | I2V models · VBench++ I2V | Background Consistency | Hurt |
| current I2V models on camera motion control | I2V models · VBench++ I2V | Camera Motion: 35.44% | No clear effect |
| industrial-developed instead of academic models for trustworthiness | T2V models · VBench++ Trustworthiness (T2V) | Culture Fairness: 84.92% vs 49.29% | Improved |
Closest papers by meaning
- VBench: Comprehensive Benchmark Suite for Video Generative Models (CVPR 2024 · similarity 0.99)
- Video-Bench: Human-Aligned Video Generation Benchmark (CVPR 2025 · similarity 0.95)
- EvalCrafter: Benchmarking and Evaluating Large Video Generation Models (CVPR 2024 · similarity 0.92)
- MovieBench: A Hierarchical Movie Level Dataset for Long Video Generation (CVPR 2025 · similarity 0.90)
- WorldModelBench: Judging Video Generation Models As World Models (NeurIPS 2025 · similarity 0.89)
- T2V-CompBench: A Comprehensive Benchmark for Compositional Text-to-video Generation (CVPR 2025 · similarity 0.89)
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.
