Dimensions overall score 8.0

• PackForcing achieves a 32x token reduction for mid tokens through massive spatiotemporal compression using a dual-branch network.technical 90%
• PackForcing can generate coherent 2-minute, 832x480 videos at 16 FPS on a single H200 GPU.result 95%
• PackForcing achieves a bounded KV cache of just 4 GB for long video generation.technical 90%
• PackForcing enables 24x temporal extrapolation (5s to 120s) for video generation.result 90%
• PackForcing achieves state-of-the-art temporal consistency (26.07) and dynamic degree (56.25) on VBench.result 95%
• PackForcing can operate effectively either zero-shot or trained on merely 5-second clips.method 85%
• PackForcing introduces a three-partition KV-cache strategy with sink, mid, and recent tokens to manage video generation history.method 90%
• The technology might face challenges in generalizing to all types of video content, especially highly complex or specialized genres.limitation 70%

Competitive landscape

Competitor map is still being generated for this paper. Enable generation or check back soon.

Startup potential card