Stability of long-horizon video generation in autoregressive diffusion models

Develop training and architectural techniques that improve the stability of long-horizon video generation in autoregressive diffusion-based models, including the distilled causal-attention version of StereoWorld, so that visual quality does not noticeably degrade as sequence length increases in both stereo and monocular settings.

Background

StereoWorld is distilled from a bidirectional attention diffusion model into a causal, autoregressive generator to enable long-horizon stereo video synthesis with a KV-cache, following Self-Forcing-style distillation. This yields significant speed-ups (from 0.49 FPS to approximately 5 FPS) and removes the 49-frame limit.

Despite these gains, the authors observe quality degradation as the generated video length increases. They note that this issue is also present in prior work (e.g., Self-Forcing), and explicitly identify stabilizing long-horizon generation as an unresolved challenge affecting both monocular and stereo video synthesis.

References

Improving the stability of long-horizon video generation therefore remains an open challenge shared by both monocular and stereo video synthesis.

— Stereo World Model: Camera-Guided Stereo Video Generation  (2603.17375 - Sun et al., 18 Mar 2026) in Supplementary Material, Section "Long Video Distillation"

While Ring Forcing significantly improves long-term memory and establishes robust object permanence in autoregressive video diffusion, minutes-long generation remains a challenging open problem.

— Ring Forcing: Towards Precise Long-Term Memory for Autoregressive Video Diffusion  (2608.26794 - Xue et al., 27 Aug 2026) in Appendix, Section 'Limitations and Future Work'

All controlled and scaling runs evaluate only 5-second videos, so we have not verified that CRP keeps suppressing error accumulation over longer autoregressive rollouts.

— Conditional Residual Prediction: Improving Autoregressive Video Diffusion without a Bidirectional Teacher  (2610.11479 - Zheng et al., 8 Oct 2026) in Conclusion, limitations paragraph