AI Paper Atlas

Research map / Video / Video generation

Quant VideoGen: Auto-Regressive Long Video Generation via 2-Bit KV-Cache Quantization

Haocheng Xi, Shuo Yang, Yilong Zhao, Muyang Li, Han Cai, Xingyang Li, Yujun Lin, Zhuoyang Zhang, Jintao Zhang, Xiuyu Li, Zhiying Xu, Jun Wu and 4 more

ICML 2026 · Video › Video generation · approach: Autoregressive long video generation · 0 citations (OpenAlex)

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 changedWhereMeasuredEffect
use QVG instead of QuaRot at INT2KV cache quantization · LongCat-Video-13B INT2PSNR: 21.573 -> 28.716Improved
use QVG instead of QuaRot at INT2KV cache quantization · HY-WorldPlay-8B INT2PSNR: 25.207 -> 29.174Improved
use QVG-Pro (4 stages, group 16) instead of QVG (ablation)QVG · LongCat-Video-13B INT2PSNR: 28.716 -> 30.376Improved
use QVG-Pro instead of QVG (ablation)QVG · LongCat-Video-13B INT2compression ratio: 6.94x -> 4.97xHurt
use QVG instead of QuaRot at INT4KV cache quantization · LongCat-Video-13B INT4PSNR: 33.744 -> 37.141Improved
use QVG instead of QuaRot at INT4KV cache quantization · HY-WorldPlay-8B INT4PSNR: 33.997 -> 34.454Improved
use QVG instead of KIVI at 1400 framesKV cache quantization · Self-Forcing 1400 framesImage Quality: 35.85 -> 67.28Improved
use QVG instead of QuaRot at 720pKV cache quantization · LongCat-Video-13B 720pPSNR: 20.88 -> 25.99Improved
add more progressive residual quantization stages (ablation)QVG · stage sweepMSE reduction: 5.83x first stage, at least 1.10x each later stageImproved
increase codebook size K from 64 to 512 (ablation)QVG INT2, 1 stage · codebook sweepkey reconstruction improvement: 4.832 -> 6.428Improved
increase codebook size K from 64 to 512 (ablation)QVG INT2, 1 stage · codebook sweepcompression ratio: 7.760 -> 7.307Hurt
increase chunk size from 9360 to 37440 (ablation)QVG Self-Forcing INT2 · chunk size sweepcompression ratio: 5.8x -> 7.0xImproved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy