# Self Gradient Forcing: Native Long Video Extrapolation

arXiv: https://arxiv.org/abs/2607.20368 · HTML: https://mlsanity.com/paper/2607.20368/

## Key facts

- Cited by 17 papers in the map (data through 14 Oct 2026).
- Placed in Video generation, one of 14 directions in Diffusion models.
- Builds on 44 of its 45 references that are in the map.
- First posted on arXiv 22 Jul 2026.

## Placed in

- Diffusion models / [Video generation](https://mlsanity.com/diffusion/video-generation/)

## Builds on (most cited first, 30 of 44)

- [Denoising Diffusion Probabilistic Models](https://mlsanity.com/paper/2006.11239/) (2020)
- [High-Resolution Image Synthesis with Latent Diffusion Models](https://mlsanity.com/paper/2112.10752/) (2021)
- [Denoising Diffusion Implicit Models](https://mlsanity.com/paper/2010.02502/) (2020)
- [Scalable Diffusion Models with Transformers](https://mlsanity.com/paper/2212.09748/) (2022)
- [RoFormer: Enhanced Transformer with Rotary Position Embedding](https://mlsanity.com/paper/2104.09864/) (2021)
- [Stable Video Diffusion: Scaling Latent Video Diffusion Models to Large Datasets](https://mlsanity.com/paper/2311.15127/) (2023)
- [Video Diffusion Models](https://mlsanity.com/paper/2204.03458/) (2022)
- [Efficient Streaming Language Models with Attention Sinks](https://mlsanity.com/paper/2309.17453/) (2023)
- [Transformer-XL: Attentive Language Models Beyond a Fixed-Length Context](https://mlsanity.com/paper/1901.02860/) (2019)
- [Imagen Video: High Definition Video Generation with Diffusion Models](https://mlsanity.com/paper/2210.02303/) (2022)
- [Make-A-Video: Text-to-Video Generation without Text-Video Data](https://mlsanity.com/paper/2209.14792/) (2022)
- [Align your Latents: High-Resolution Video Synthesis with Latent Diffusion Models](https://mlsanity.com/paper/2304.08818/) (2023)
- [AnimateDiff: Animate Your Personalized Text-to-Image Diffusion Models without Specific Tuning](https://mlsanity.com/paper/2307.04725/) (2023)
- [SnapKV: LLM Knows What You are Looking for Before Generation](https://mlsanity.com/paper/2404.14469/) (2024)
- [Diffusion Forcing: Next-token Prediction Meets Full-Sequence Diffusion](https://mlsanity.com/paper/2407.01392/) (2024)
- [Movie Gen: A Cast of Media Foundation Models](https://mlsanity.com/paper/2410.13720/) (2024)
- [Self Forcing: Bridging the Train-Test Gap in Autoregressive Video Diffusion](https://mlsanity.com/paper/2506.08009/) (2025)
- [VideoCrafter2: Overcoming Data Limitations for High-Quality Video Diffusion Models](https://mlsanity.com/paper/2401.09047/) (2024)
- [VideoPoet: A Large Language Model for Zero-Shot Video Generation](https://mlsanity.com/paper/2312.14125/) (2023)
- [Lumiere: A Space-Time Diffusion Model for Video Generation](https://mlsanity.com/paper/2401.12945/) (2024)
- [From Slow Bidirectional to Fast Autoregressive Video Diffusion Models](https://mlsanity.com/paper/2412.07772/) (2024)
- [LongLive: Real-time Interactive Long Video Generation](https://mlsanity.com/paper/2509.22622/) (2025)
- [StreamingT2V: Consistent, Dynamic, and Extendable Long Video Generation from Text](https://mlsanity.com/paper/2403.14773/) (2024)
- [Rolling Forcing: Autoregressive Long Video Diffusion in Real Time](https://mlsanity.com/paper/2509.25161/) (2025)
- [Self-Forcing++: Towards Minute-Scale High-Quality Video Generation](https://mlsanity.com/paper/2510.02283/) (2025)
- [Causal Forcing: Autoregressive Diffusion Distillation Done Right for High-Quality Real-Time Interactive Video Generation](https://mlsanity.com/paper/2602.02214/) (2026)
- [FreeNoise: Tuning-Free Longer Video Diffusion via Noise Rescheduling](https://mlsanity.com/paper/2310.15169/) (2023)
- [Gen-L-Video: Multi-Text to Long Video Generation via Temporal Co-Denoising](https://mlsanity.com/paper/2305.18264/) (2023)
- [FIFO-Diffusion: Generating Infinite Videos from Text without Training](https://mlsanity.com/paper/2405.11473/) (2024)
- [Matrix-Game 3.0: Real-Time and Streaming Interactive World Model with Long-Horizon Memory](https://mlsanity.com/paper/2604.08995/) (2026)

## Built on by (most cited first, 17 of 17)

- [EchoWM: Open and Enterable Omnimodal World Models](https://mlsanity.com/paper/2608.23189/) (2026)
- [Matrix-Game 3.5: Enhancing Real-Time Streaming Interactive World Models with Patch Memory](https://mlsanity.com/paper/2608.29910/) (2026)
- [Context-Matched Distillation: Teacher Causality for Autoregressive Video Distillation](https://mlsanity.com/paper/2608.13391/) (2026)
- [Long-Horizon Audio-Visual Generation for Persistent Stories and Interactive Worlds](https://mlsanity.com/paper/2608.23383/) (2026)
- [Zing-0.5: Toward Playable Worlds with Real-Time Joint Action and Text Control](https://mlsanity.com/paper/2609.17909/) (2026)
- [RECAP-Forcing: Retaining Content Appearances for Long Video Generation](https://mlsanity.com/paper/2608.26671/) (2026)
- [XPACE: Joint World and Action Modeling from Heterogeneous Experience](https://mlsanity.com/paper/2609.17372/) (2026)
- [Precise Editing and Flexible Referencing for Interactable Worlds](https://mlsanity.com/paper/2609.34470/) (2026)
- [ViRDM: Taming Representation Distribution Matching for Few-Step Causal Video Generation](https://mlsanity.com/paper/2609.28923/) (2026)
- [Triangular Resampling for Long-Horizon Motion Generation](https://mlsanity.com/paper/2609.34697/) (2026)
- [From Scores to Samples: Elastic Forcing for Autoregressive Video Generation](https://mlsanity.com/paper/2609.35491/) (2026)
- [Self-Aligned Forcing: Streaming Video Diffusion with Differentiable Noisy History](https://mlsanity.com/paper/2609.38114/) (2026)
- [LongTake: Learning to Sustain Dynamics in Long-Horizon Video Generation](https://mlsanity.com/paper/2609.38562/) (2026)
- [SGF+: Decoupling Gradient Flows for Autoregressive Video Generation](https://mlsanity.com/paper/2610.10429/) (2026)
- [Connected Self Forcing: Beyond Local Learning in Video Autoregression](https://mlsanity.com/paper/2610.12156/) (2026)
- [AgentGarten: Code Worlds for Evolving Agents](https://mlsanity.com/paper/2610.12374/) (2026)
- [Adversarial Gradient Forcing for Robust Few-Step Autoregressive Video Generation](https://mlsanity.com/paper/2610.14553/) (2026)

## Abstract (arXiv)

Recent autoregressive video diffusion methods are increasingly built upon Self Forcing, where the student is trained on histories produced by its own rollout rather than ground-truth video contexts. This reduces exposure bias, but the historical key-value cache is still used by future frames only as frozen rollout state. As a result, future losses cannot supervise how earlier generated latents should be written into more useful keys and values for later video-latent generation. We call this the historical context-gradient gap. We propose Self Gradient Forcing (SGF), a two-pass training strategy that restores this missing supervision signal without backpropagating through the full serial rollout. Pass 1 performs a no-gradient autoregressive rollout matching inference and, at a sampled denoising exit step, records both the self-generated context and the noisy latents fed to the model. Pass 2 performs parallel context-gradient reconstruction for the recorded exit step. The generated context is used as stop-gradient clean-latent input, while the model recomputes the context KV representations and future-to-context causal attention. Thus, SGF provides the missing memory-writing supervision within the native autoregressive training objective, using losses on future video latents to train the model to encode context into more effective causal memory. Across extensive long-horizon frame-wise and chunk-wise experiments under different initializations, SGF achieves stronger native long-video extrapolation than Self Forcing, especially in subject identity, background/layout consistency, and temporal stability. Remarkably, using only a 5-second training window, SGF can extrapolate to videos lasting several minutes. Code and models will be released to advance research on autoregressive video generation.