Research map / Video / Video generation
Ref4D-VideoBench: Four-Dimensional Reference-Based Evaluation of Text-to-Video Generative Models
Jiajia Wei, Yujia He, Yuhan Hou, Hang Qi, Sihua Wang, Jincheng Shi, Kwok Fung Li, Zibin Zheng, Weibin 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 |
|---|---|---|---|
| add per-video rule/question bank (ablation) | MiniCPM-V-4.5 world-knowledge scorer · World knowledge consistency vs human | SRCC: 0.413 -> 0.847 | Improved |
| add per-video rule/question bank (ablation) | VideoLLaMA3-7B world-knowledge scorer · World knowledge consistency vs human | SRCC: 0.406 -> 0.757 | Improved |
| use MiniCPM-V-4.5 instead of VideoLLaMA3-7B as judge (ablation) | world-knowledge scorer · World knowledge consistency vs human | SRCC: 0.757 -> 0.847 | Improved |
| use reference-based framework instead of best no-reference baseline | Ref4D · Human alignment: Semantic | SRCC: 0.317 -> 0.822 | Improved |
| use reference-based framework instead of best baseline | Ref4D · Human alignment: Motion | SRCC: 0.358 -> 0.659 | Improved |
| use reference-based framework instead of best baseline | Ref4D · Human alignment: Event | SRCC: 0.220 -> 0.755 | Improved |
| use reference-based framework instead of best baseline | Ref4D · Human alignment: World knowledge | SRCC: 0.391 -> 0.847 | Improved |
| use learned linear aggregator over atomic metrics (semantic) (ablation) | Ref4D · Human alignment: Semantic | SRCC: 0.734 -> 0.822 | Improved |
| use learned linear aggregator over atomic metrics (event) (ablation) | Ref4D · Human alignment: Event | SRCC: 0.718 -> 0.755 | Improved |
| use learned linear aggregator over atomic metrics (motion) (ablation) | Ref4D · Human alignment: Motion | SRCC: -0.498 (Dmag) -> 0.659 | Improved |
Closest papers by meaning
- VBench: Comprehensive Benchmark Suite for Video Generative Models (CVPR 2024 · similarity 0.91)
- Video-Bench: Human-Aligned Video Generation Benchmark (CVPR 2025 · similarity 0.91)
- EvalCrafter: Benchmarking and Evaluating Large Video Generation Models (CVPR 2024 · similarity 0.89)
- T2V-CompBench: A Comprehensive Benchmark for Compositional Text-to-video Generation (CVPR 2025 · similarity 0.89)
- MMBench-Video: A Long-Form Multi-Shot Benchmark for Holistic Video Understanding (NeurIPS 2024 · similarity 0.89)
- Neuro-Symbolic Evaluation of Text-to-Video Models using Formal Verification (CVPR 2025 · similarity 0.88)
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.
