AI Paper Atlas

Research map / Video / Video generation

Ring Forcing: Towards Precise Long-Term Memory for Autoregressive Video Diffusion

Bowen Xue, Brandon Feng, Chenguo Lin, Yuchen Lin, Yujia Zeng, lvmin zhang, Maneesh Agrawala, Honglei Yan, Panwang Pan

ECCV 2026 · Video › Video generation · approach: Autoregressive long video generation · 8 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 sparse RoPE mapping compressed tokens to physical coordinates instead of index-based RoPE (ablation)Ring Forcing · reconstructionPSNR: 19.05 -> 24.22Improved
use sparse RoPE instead of standard RoPE (ablation)Ring Forcing · reconstructionLPIPS: 0.27 -> 0.13Improved
increase ring context drop probability from 0.0 to 1.0 (ablation)Ring Forcing · ring-mode reconstructionPSNR: 25.02 -> 21.86Hurt
increase ring context drop probability from 0.0 to 1.0 (ablation)Ring Forcing · standard-mode diversityDiversity Score: 3.71 -> 4.23Improved
choose p_drop=0.4 over 0.0 (ablation)Ring Forcing · reconstruction vs diversityPSNR / Diversity: PSNR 25.02 -> 24.22; Diversity 3.71 -> 4.11Mixed
timestep composition (global + detail streams) vs single-stream baselines (ablation)Ring Forcing · reconstruction SSIMSSIMImproved
global stream only (high spatial compression) vs detail stream only (aggressive temporal compression) (ablation)Ring Forcing · reconstruction SSIMSSIMImproved
ring-structured training vs linear-trained baselines at 1s gapRing Forcing · A-D-R benchmarkTexture consistency: 0.252 -> 0.723Improved
ring-structured training vs baselines at 1s gapRing Forcing · A-D-R benchmarkGeometry consistency: 0.368 -> 0.839Improved
Ring Forcing vs base model on aestheticsRing Forcing · General benchmarkAesthetic: 5.825 -> 5.822No clear effect
longer disappearance gap (15s to 60s)Ring Forcing · A-D-R benchmarkTexture consistency: 0.509 -> 0.396Hurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most 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

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