AIAny
Icon for item

Self Gradient Forcing: Native Long Video Extrapolation

Extrapolates long video sequences from very short contexts by restoring memory-writing supervision in autoregressive video diffusion models using a two-pass Self Gradient Forcing (SGF). SGF records a no-gradient rollout at a sampled denoising exit and then recomputes KV context in a second parallel pass so future losses teach earlier latent writes, enabling minutes-long extrapolation from ~5s windows.

Introduction

Most autoregressive video diffusion pipelines train on ground-truth contexts or on their own rollouts but still treat the historical key–value cache as a frozen past, leaving a gap: future losses cannot teach earlier generations how to write more useful context for later frames. That missing supervision—what the authors call the "historical context-gradient gap"—is why many native autoregressive generators struggle to preserve subject identity, scene layout, and temporal stability when extrapolating far beyond their training window.

Key Findings
  • Restoring missing supervision without full backpropagation: SGF runs a no-gradient autoregressive rollout that mimics inference and records the generated context plus noisy latents at a sampled denoising exit; a second parallel pass recomputes KV representations and causal attention so future-to-context losses can update how earlier latents are written. This gives memory-writing supervision while avoiding the compute of full serial backprop.
  • Large extrapolation from tiny windows: In experiments, SGF trained on very short (≈5s) windows can extrapolate video continuations lasting minutes, improving subject identity preservation, background/layout consistency, and temporal stability compared to standard Self Forcing.
  • Training- and deployment-friendly: The two-pass design matches inference rollout behavior in the first pass (reducing train–inference mismatch) while keeping the heavy gradient computation localized to sampled exits, making it more practical to integrate into existing autoregressive video diffusion setups.
Who it's for and tradeoffs

Great fit if you develop or research autoregressive video generation and need better long-horizon extrapolation from limited context—especially when preserving identity and scene consistency matters. Look elsewhere if you require an out-of-the-box, zero-infrastructure solution: SGF adds complexity (a two-pass training loop and exit-step bookkeeping) and requires modifications to model training and memory/cache handling. It also focuses on improving native autoregressive objectives rather than substituting alternative architectures (e.g., retrieval-conditioned or latent diffusion approaches) that might be preferable for some production constraints.

Method note

The core trick is decoupling the forward rollout (no gradients) from a targeted gradient-aware reconstruction at selected denoising exits. By recomputing context KV pairs and causal attention in parallel for those exits, SGF channels future-latent losses back into how the model writes context, effectively teaching the model to create more useful causal memory without the prohibitive cost of end-to-end rollout backpropagation.

Information

  • Websitearxiv.org
  • AuthorsJunhao Zhuang, Shiyi Zhang, Yuxuan Bian, Yaowei Li, Yawen Luo, Yijun Liu, Weiyang Jin, Songchun Zhang, Xianglong He, Xuying Zhang
  • Published date2026/07/22

More Items

Quantifies active visual observation in multimodal LLMs with ActiveVision, a 17-task benchmark that forces repeated perception rather than one-shot description. Finds frontier MLLMs fail badly (top model 10.6% vs humans 96.1%) and that model-generated vision code does not close the gap.

Generates interactive long-horizon 24-fps video worlds (540p/720p) from text, image, or video inputs. Uses a 15B video diffusion transformer with a bounded visual context (sink frame, compressed temporal history, geometry-aligned spatial memory, recent-frame conditioning) and a discrete autoregressive distillation that cuts inference to ~4 sampling steps per chunk.

Efficient 4B-scale image generation and editing model family that pairs a lightweight VAE tokenizer (Mage-VAE) with a native-resolution multimodal diffusion transformer, reducing tokenization cost by an order of magnitude and enabling few-step high-resolution generation and editing.