Research map / Video / Video-language understanding
VideoCoF: Unified Video Editing with Temporal Reasoner
Xiangpeng Yang, Ji Xie, Yiyuan Yang, Yue Ma, Yan Huang, Min Xu, Qiang 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 chain-of-frames reasoning frames with RoPE [1-F,0,1-F] instead of naive temporal in-context (ablation) | VideoCoF Wan-14B · VideoCoF-Bench | Instruct Follow (GPT-4o): 8.109 -> 8.973 | Improved |
| add chain-of-frames reasoning frames with RoPE [1-F,0,1-F] instead of naive temporal in-context (ablation) | VideoCoF Wan-14B · VideoCoF-Bench | Success Ratio: 72.41% -> 76.36% | Improved |
| add chain-of-frames reasoning frames with RoPE [1-F,0,1-F] instead of naive temporal in-context (ablation) | VideoCoF Wan-14B · VideoCoF-Bench | CLIP-T: 26.880 -> 28.000 | Improved |
| set reasoning frame RoPE index to 0 and source/target to 1..F instead of 0-F-1 for all (ablation) | VideoCoF Wan-14B · VideoCoF-Bench | Success Ratio: 65.52% -> 76.36% | Improved |
| reset RoPE indices to avoid index collision (repeat indices) rather than sequential 0-2F-1 (ablation) | VideoCoF Wan-14B · VideoCoF-Bench | length extrapolation to 81+ frames | Improved |
| use reasoning frame index 0 colliding with first frame index 0 (ablation) | VideoCoF Wan-14B · VideoCoF-Bench | Success Ratio: 76.36% -> 65.52% | Hurt |
| use gray 50% mask instead of black mask for reasoning frame (ablation) | VideoCoF Wan-14B · VideoCoF-Bench | Success Ratio: 52.17% -> 68.45% | Improved |
| use gray 50% mask instead of red 50% highlight (ablation) | VideoCoF Wan-14B · VideoCoF-Bench | Success Ratio: 60.33% -> 68.45% | Improved |
| use progressive gray transparency 0-75% instead of static 50% gray (ablation) | VideoCoF Wan-14B · VideoCoF-Bench | Success Ratio: 68.45% -> 76.36% | Improved |
| use progressive gray transparency 0-75% instead of static 50% gray (ablation) | VideoCoF Wan-14B · VideoCoF-Bench | Instruct Follow (GPT-4o): 8.150 -> 8.973 | Improved |
| use VideoCoF with 50k pairs versus ICVE trained on 1M plus 150k | VideoCoF vs ICVE · VideoCoF-Bench | Success Ratio: 57.76% -> 76.36% | Improved |
| use VideoCoF versus ICVE | VideoCoF vs ICVE · VideoCoF-Bench | Quality (GPT-4o): 8.14 -> 7.77 | Hurt |
Later papers that compared with this method
2 later accepted papers report VideoCoF as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- OpenVE-3M: A Large-Scale High-Quality Dataset for Instruction-Based Video Editing (ECCV 2026 · better on 2 of 3 shared results)
- LiveEdit: Towards Real-Time Diffusion-Based Streaming Video Editing (ECCV 2026 · better on 3 of 3 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- OpenCoF: Learning to Reason Through Video Generation (NeurIPS 2026 · similarity 0.88)
- VEU-Bench: Towards Comprehensive Understanding of Video Editing (CVPR 2025 · similarity 0.87)
- VINCIE-NExT: Unlocking Video Editing from Images via In-Context Modeling (NeurIPS 2026 · similarity 0.87)
- Learning Action and Reasoning-Centric Image Editing from Videos and Simulation (NeurIPS 2024 · similarity 0.86)
- Señorita-2M: A High-Quality Instruction-based Dataset for General Video Editing by Video Specialists (NeurIPS 2025 · similarity 0.86)
- VicEdit: Learning to Edit Videos from Visual In-Context Examples (NeurIPS 2026 · similarity 0.85)
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.
