Research map / Video / Video generation
NarrLV: Towards a Comprehensive Narrative-Centric Evaluation for Long Video Generation
Xiaokun Feng, Haiming Yu, Meiqi Wu, Shiyu Hu, Jintao Chen, Chen Zhu, Jiahong Wu, Xiangxiang Chu, Kaiqi Huang
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 more Temporal Narrative Atoms to the prompt | foundation and long video models · NarrLV | R_cov and R_coh | Hurt |
| add more Temporal Narrative Atoms to the prompt | foundation and long video models · NarrLV | R_fid | No clear effect |
| add long-video modules (FIFO-Diffusion, FreeLong, FreePCA, FreeNoise) on VideoCraft | VideoCraft · NarrLV | R_cov and R_coh | Improved |
| use object action as the TNA change factor instead of scene or object attributes | average over models · NarrLV | R_cov / R_coh: R_cov 53.7 vs 57.4 and 60.3; R_coh 37.9 vs 39.6 and 40.7 | Hurt |
| use object action as the TNA change factor | average over models · NarrLV | R_fid: 71.4 vs 67.9 and 67.6 | Improved |
| ask the MLLM to answer 5 times instead of 1 or 3 (ablation) | NarrLV metric · Human alignment accuracy | alignment accuracy | Improved |
| use Qwen2.5-VL-72B instead of Qwen2.5-VL-32B as the MLLM (ablation) | NarrLV metric · Human alignment accuracy | alignment accuracy | Improved |
| increase MLLM sample count from 1 to 5 (ablation) | NarrLV metric · Reproducibility | mean absolute error (Wan): 0.0031 -> 0.0009 | Improved |
Closest papers by meaning
- Is Your World Simulator a Good Story Presenter? A Consecutive Events-Based Benchmark for Future Long Video Generation (CVPR 2025 · similarity 0.89)
- MovieBench: A Hierarchical Movie Level Dataset for Long Video Generation (CVPR 2025 · similarity 0.88)
- Evaluation of Text-to-Video Generation Models: A Dynamics Perspective (NeurIPS 2024 · similarity 0.87)
- Advancing Narrative Long Video Generation via Training-Free Identity-Aware Memory (NeurIPS 2026 · similarity 0.87)
- MovieDreamer: Hierarchical Generation for Coherent Long Visual Sequences (ICLR 2025 · similarity 0.87)
- LVD-2M: A Long-take Video Dataset with Temporally Dense Captions (NeurIPS 2024 · similarity 0.86)
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.
