Research map / Video / Video generation
VideoRealBench: A Chain-of-Thought Realism Evaluation Benchmark for Generated Human-Centric Videos
Min Yang, Xinwen Zhang, Jialei Tang, Xin Zhou, Kehan Li, Zeyi Huang, Limin Wang
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 |
|---|---|---|---|
| fine-tune Qwen2.5-VL-7B with LoRA on VideoRealDataset | Qwen2.5-VL-7B · VideoRealDataset test | PLCC (%): 18.93 -> 57.07 | Improved |
| add Problem Description step to CoT (ablation) | VideoRealEval · VideoRealDataset test | PLCC (%): 53.89 -> 55.99 | Improved |
| add Standard Adherence step to CoT (ablation) | VideoRealEval · VideoRealDataset test | PLCC (%): 55.99 -> 57.07 | Improved |
| use integer scores 1-5 instead of verbal score words (ablation) | VideoRealEval · VideoRealDataset test | PLCC (%): 56.89 -> 57.07 | No clear effect |
| re-annotate scores with explicit criteria instead of original VideoPhy-2 labels (ablation) | VideoRealEval · VideoRealDataset test | PLCC (%): 41.87 -> 53.89 | Improved |
| add re-annotated CoT rationales on top of re-annotated scores (ablation) | VideoRealEval · VideoRealDataset test | PLCC (%): 53.89 -> 57.07 | Improved |
| use VideoRealBench re-annotated scores instead of VideoPhy-2 scores | annotation · 345 video pairs | Preference Consistency Ratio: 0.754 -> 0.925 | Improved |
| train VideoPhy-2-AutoEval on VideoRealDataset | VideoPhy-2-AutoEval · VideoRealDataset test | PLCC (%): 38.31 -> 50.89 | Improved |
| zero-shot transfer of VideoRealEval versus GPT-4o on Video-Bench Motion Effects | VideoRealEval · Video-Bench Motion Effects | SROCC: 0.514 -> 0.508 | No clear effect |
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 0.90)
- EvalCrafter: Benchmarking and Evaluating Large Video Generation Models (CVPR 2024 · similarity 0.88)
- 3DSPA: A 3D Semantic Point Autoencoder for Evaluating Video Realism (NeurIPS 2026 · similarity 0.87)
- Impossible Videos (ICML 2025 · similarity 0.87)
- VideoPhy: Evaluating Physical Commonsense for Video Generation (ICLR 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.
