Research map / Video / Video generation
LoL: Longer than Longer, Scaling Video Generation to Hour
Justin Cui, Jie Wu, Ming Li, Tao Yang, Xiaojie Li, Rui Wang, Andrew Bai, Yuanhao Ban, Cho-Jui Hsieh
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 multi-head RoPE jitter (LoL) to position extrapolation | LongLive · VBench 100s videos (128 MovieGen prompts) | Sink-Collapse Max: 73.06 -> 16.67 | Improved |
| add multi-head RoPE jitter (LoL) to position extrapolation | LongLive · VBench 100s videos (128 MovieGen prompts) | Sink-Collapse Avg: 30.54 -> 3.93 | Improved |
| add multi-head RoPE jitter (LoL) to position extrapolation | Self-Forcing++ · VBench 100s videos (128 MovieGen prompts) | Sink-Collapse Max: 68.07 -> 22.70 | Improved |
| add multi-head RoPE jitter (LoL) to position extrapolation | LongLive · VBench 100s videos (128 MovieGen prompts) | Dynamic Degree: 34.62 -> 35.27 | No clear effect |
| add multi-head RoPE jitter (LoL) to position extrapolation | Self-Forcing++ · VBench 100s videos (128 MovieGen prompts) | Dynamic Degree: 83.32 -> 81.20 | No clear effect |
| use position interpolation instead of extrapolation | LongLive · VBench 100s videos (128 MovieGen prompts) | Dynamic Degree: 34.62 -> 00.35 | Hurt |
| use YARN instead of extrapolation | LongLive · VBench 100s videos (128 MovieGen prompts) | Dynamic Degree: 34.62 -> 2.67 | Hurt |
| use NTK instead of extrapolation | LongLive · VBench 100s videos (128 MovieGen prompts) | Sink-Collapse Max: 73.06 -> 41.11 | Mixed |
| use RIFLEx single-dimension frequency change instead of extrapolation (ablation) | LongLive · VBench 100s videos (128 MovieGen prompts) | Sink-Collapse Max: 73.06 -> 70.95 | No clear effect |
| vary RoPE base theta from 6000 to 20000 (ablation) | LongLive · RoPE base sweep | sink-collapse position | No clear effect |
| alter frequency of any single RoPE dimension (ablation) | LongLive · RoPE dimension sweep | sink-collapse | No clear effect |
| increase jitter intensity sigma from 0.1 to 0.8 (ablation) | LongLive · Jitter intensity sweep | sink-collapse: 0.1 -> 0.8 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- RIFLEx — RIFLEx: A Free Lunch for Length Extrapolation in Video Diffusion Transformers (ICML 2025) · better on 5 of 6 shared results
- LongLive — LongLive: Real-time Interactive Long Video Generation (ICLR 2026) · better on 1 of 1 shared result
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
- WorldWeaver: Generating Long-Horizon Video Worlds via Rich Perception (NeurIPS 2025 · similarity 0.87)
- Generative Inbetweening through Frame-wise Conditions-Driven Video Generation (CVPR 2025 · similarity 0.87)
- AdaState: Self-Evolving Anchors for Streaming Video Generation (NeurIPS 2026 · similarity 0.86)
- StreamingT2V: Consistent, Dynamic, and Extendable Long Video Generation from Text (CVPR 2025 · similarity 0.86)
- LongDiff: Training-Free Long Video Generation in One Go (CVPR 2025 · similarity 0.86)
- FreeLong: Training-Free Long Video Generation with SpectralBlend Temporal Attention (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.
