Ca2-VDM: Efficient Autoregressive Video Diffusion Model with Causal Generation and Cache Sharing
Reviewed by Pith T0 review T1 audit T2 compute T3 formal T4 kernel pith:LWBJGY3Urecord.jsonopen to challenge →
read the original abstract
With the advance of diffusion models, today's video generation has achieved impressive quality. To extend the generation length and facilitate real-world applications, a majority of video diffusion models (VDMs) generate videos in an autoregressive manner, i.e., generating subsequent clips conditioned on the last frame(s) of the previous clip. However, existing autoregressive VDMs are highly inefficient and redundant: The model must re-compute all the conditional frames that are overlapped between adjacent clips. This issue is exacerbated when the conditional frames are extended autoregressively to provide the model with long-term context. In such cases, the computational demands increase significantly (i.e., with a quadratic complexity w.r.t. the autoregression step). In this paper, we propose Ca2-VDM, an efficient autoregressive VDM with Causal generation and Cache sharing. For causal generation, it introduces unidirectional feature computation, which ensures that the cache of conditional frames can be precomputed in previous autoregression steps and reused in every subsequent step, eliminating redundant computations. For cache sharing, it shares the cache across all denoising steps to avoid the huge cache storage cost. Extensive experiments demonstrated that our Ca2-VDM achieves state-of-the-art quantitative and qualitative video generation results and significantly improves the generation speed. Code is available: https://github.com/Dawn-LX/CausalCache-VDM
This paper has not been read by Pith yet.
Forward citations
Cited by 21 Pith papers
-
MV-Forcing: Long Multi-View Video Generation via 4D-Grounded Spatio-Temporal Self-Forcing
MV-Forcing composes temporal and view-sequential autoregression in a single diffusion model, using a recurrent 3D reconstruction model as a geometric bridge to generate arbitrarily long, multi-view consistent videos.
-
TempAct: Advancing Temporal Plausibility in Autoregressive Video Generation via Planner-Executor RL
TempAct introduces a planner-executor RL framework with hierarchical group exploration and rewards to improve temporal consistency in autoregressive video diffusion models.
-
HL-OutPaint: Coarse-to-Fine Video Outpainting for High-Resolution Long-Range Videos
HL-OutPaint enables high-resolution outpainting of long video sequences via a coarse-to-fine pipeline that first builds Global Coarse Guidance through global-local frame swapping then synthesizes details.
-
Efficient Video Diffusion Models: Advancements and Challenges
A survey that groups efficient video diffusion methods into four paradigms—step distillation, efficient attention, model compression, and cache/trajectory optimization—and outlines open challenges for practical use.
-
ChopGrad: Pixel-Wise Losses for Latent Video Diffusion via Truncated Backpropagation
ChopGrad truncates backpropagation to local frame windows in video diffusion models, reducing memory from linear in frame count to constant while enabling pixel-wise loss fine-tuning.
-
Latent Generative Solvers for Generalizable Long-Term Physics Simulation
LGS pretrained on 2.5M trajectories across 16 systems matches deterministic baselines at one step and halves 20-step error while using far less compute and adapting to held-out higher-resolution flows.
-
RoboWorld: Fast and Reliable Neural Simulators for Generalist Robot Policy Evaluation
RoboWorld introduces an automated pipeline using autoregressive video world models and task-progress VLM scoring, plus Step Forcing for long-horizon stability, to achieve high correlation with real robot policy evaluation.
-
TempAct: Advancing Temporal Plausibility in Autoregressive Video Generation via Planner-Executor RL
TempAct applies hierarchical planner-executor RL with group exploration and multi-level rewards to improve temporal consistency in autoregressive video models.
-
MORPHOS: Autoregressive 4D Generation with Temporal Structured Latents
MORPHOS introduces an autoregressive 4D generation method with Temporal Structured Latents (T-SLAT) that produces dynamic 3D assets from videos while handling topological changes and long sequences.
-
HL-OutPaint: Coarse-to-Fine Video Outpainting for High-Resolution Long-Range Videos
HL-OutPaint builds a global coarse guidance representation via global-local frame swapping to guide high-resolution outpainting for long-range videos.
-
FashionChameleon: Towards Real-Time and Interactive Human-Garment Video Customization
FashionChameleon achieves interactive multi-garment video customization in real time by training a teacher model with in-context learning on single-garment pairs, applying streaming distillation, and using training-fr...
-
FashionChameleon: Towards Real-Time and Interactive Human-Garment Video Customization
FashionChameleon achieves interactive multi-garment video customization at 23.8 FPS via in-context teacher models, streaming distillation, and training-free KV cache rescheduling while using only single-garment data.
-
Head Forcing: Long Autoregressive Video Generation via Head Heterogeneity
Head Forcing assigns tailored KV cache strategies to local, anchor, and memory attention heads plus head-wise RoPE re-encoding to extend autoregressive video generation from seconds to minutes without training.
-
Stream-T1: Test-Time Scaling for Streaming Video Generation
Stream-T1 is a test-time scaling framework for streaming video generation using scaled noise propagation from history, reward pruning across short and long windows, and feedback-guided memory sinking to improve tempor...
-
INSPATIO-WORLD: A Real-Time 4D World Simulator via Spatiotemporal Autoregressive Modeling
INSPATIO-WORLD is a real-time framework for high-fidelity 4D scene generation and navigation from monocular videos via STAR architecture with implicit caching, explicit geometric constraints, and distribution-matching...
-
World Action Models are Zero-shot Policies
DreamZero uses a 14B video diffusion model as a World Action Model to achieve over 2x better zero-shot generalization on real robots than state-of-the-art VLAs, real-time 7Hz closed-loop control, and cross-embodiment ...
-
Rolling Sink: Bridging Limited-Horizon Training and Open-Ended Testing in Autoregressive Video Diffusion
Rolling Sink is a training-free cache adjustment technique that maintains visual consistency in autoregressive video diffusion models for ultra-long open-ended generation beyond training horizons.
-
Flow marching for a generative PDE foundation model
Flow Marching jointly samples noise and physical time to learn a velocity field for generative PDE modeling, paired with a latent autoencoder and efficient transformer for large-scale pretraining on 2.5M trajectories.
-
Self Forcing: Bridging the Train-Test Gap in Autoregressive Video Diffusion
Self Forcing trains autoregressive video diffusion models by performing autoregressive rollout with KV caching during training to close the exposure bias gap, using a holistic video-level loss and few-step diffusion f...
-
Directing the World: Fast Autoregressive Video Generation with Compositional Human-Camera Control
A decoupled-control autoregressive video model using Fast-Slow Memory training, dynamic projection, and staged camera control to produce stable long-horizon outputs with human and viewpoint guidance.
-
DreamForge-World 0.1 Preview: A Low-Compute Real-Time Controllable World Model
A preview system demonstrates real-time controllable world modeling at 14-15 FPS on RTX 4090 by adapting open video backbones with action pathways for keyboard/mouse control and multimodal features.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.