Dynamic-in-Few-Step: Unifying Dynamic Computation and Few-Step Distillation for Efficient Video Generation

• URL: http://arxiv.org/abs/2607.06631v1
• Date: Tue, 07 Jul 2026 13:14:33 GMT
• Title: Dynamic-in-Few-Step: Unifying Dynamic Computation and Few-Step Distillation for Efficient Video Generation
• Abstract summary: Video Diffusion Models (VDMs) have demonstrated superior generation quality but suffer from prohibitive computational costs.<n>We propose a novel post-training acceleration framework that exploits this redundancy by integrating dynamic structural sparsification directly into the distillation process.
• Score: 24.896148135296198
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Video Diffusion Models (VDMs) have demonstrated superior generation quality but suffer from prohibitive computational costs. While recent few-step distillation techniques significantly accelerate inference, they typically enforce a static model architecture across all denoising stages, ignoring the varying computational demands inherent to different noise levels. In this work, we propose a novel post-training acceleration framework that exploits this redundancy by integrating dynamic structural sparsification directly into the distillation process. Unlike conventional post-hoc compression applied to a fixed diffusion pipeline, our approach jointly optimizes the denoising steps and structured model sparsity, transforming a pre-trained VDM into a compact, step-specific Mixture-of-Models (MoM). To address the training instability arising from this joint optimization, we introduce a Progressive Training Strategy coupled with an Output Rollout Mechanism, which ensures the coherent learning of structural decisions across timesteps. Furthermore, we develop a specialized inference engine to deploy the resulting MoM efficiently. Our method is orthogonal to existing acceleration techniques and highly effective: On Wan-14B, it removes 24% of the per-step FLOPs on top of 4-step distillation, adding a 1.2x wall-clock gain and reaching a 30x speedup over the 50-step teacher while preserving competitive generation quality.

Related papers

• DiffusionGemma Technical Report [72.10570774996219]
  We introduce DiffusionGemma, a language model that uses discrete diffusion to generate text at exceptionally high speed.<n>Rather than decoding one token at a time, DiffusionGemma iteratively refines blocks of 256 tokens in parallel.<n>Our compute-efficient two-stage training pipeline uses fewer than 10% of the starting AR model's total training token budget.
  arXiv  Detail & Related papers  (2026-07-31T16:11:46Z)
• SATB-VR: Training Few-Step Video Restoration Diffusion Model using SNR-Aware Trajectory Blending [55.764018145256216]
  We present SATB-VR, a few-step paradigm that jump-starts the denoising process via an auxiliary predictor.<n>We show that SATB-VR performs favorably against existing approaches on synthetic, real-world, and AIGC benchmarks.
  arXiv  Detail & Related papers  (2026-06-27T01:32:16Z)
• Curvature-Adaptive Consistency Flow Matching: Autonomous Trajectory Optimization via Reinforcement Learning [8.422673935994451]
  consistency distillation has significantly accelerated the inference of diffusion models.<n>We propose Curvature-Adaptive Consistency Flow Matching to address the limitations of static sampling.<n>Our approach achieves new state-of-the-art results on large-scale models such as FLUX and SDXL.
  arXiv  Detail & Related papers  (2026-06-21T08:49:50Z)
• teasr: training-efficient any-step diffusion transformer for real-world image super-resolution [10.733502031936958]
  TEASR is a training-efficient any-step diffusion framework for Real-ISR.<n>Our key idea is to perform self-adversarial distillation within a single diffusion model.<n>We propose a timestep-aware rectification strategy that stabilizes one-step generation across noise levels.
  arXiv  Detail & Related papers  (2026-06-15T04:02:37Z)
• Towards One-step Causal Video Generation via Adversarial Self-Distillation [71.30373662465648]
  Recent hybrid video generation models combine autoregressive temporal dynamics with diffusion-based spatial denoising.<n>Our framework produces a single distilled model that flexibly supports multiple inference-step settings.
  arXiv  Detail & Related papers  (2025-11-03T10:12:47Z)
• Fewer Denoising Steps or Cheaper Per-Step Inference: Towards Compute-Optimal Diffusion Model Deployment [14.097906894386066]
  PostDiff is a training-free framework for accelerating pre-trained diffusion models.<n>We show that PostDiff can significantly improve the fidelity-efficiency trade-off of state-of-the-art diffusion models.
  arXiv  Detail & Related papers  (2025-08-08T09:29:37Z)
• Training-Free Motion Customization for Distilled Video Generators with Adaptive Test-Time Distillation [53.877572078307935]
  Distilled video generation models offer fast and efficient but struggle with motion customization when guided by reference videos.<n>We propose MotionEcho, a training-free test-time distillation framework that enables motion customization by leveraging diffusion teacher forcing.
  arXiv  Detail & Related papers  (2025-06-24T06:20:15Z)
• Minitron-SSM: Efficient Hybrid Language Model Compression through Group-Aware SSM Pruning [76.88243649182886]
  Hybrid architectures that combine Attention and State Space Models (SSMs) achieve state-of-the-art accuracy and runtime performance.<n>Recent work has demonstrated that applying compression and distillation to Attention-only models yields smaller, more accurate models at a fraction of the training cost.<n>We introduce a novel group-aware pruning strategy that preserves the structural integrity of SSM blocks and their sequence modeling capabilities.
  arXiv  Detail & Related papers  (2025-04-15T17:26:29Z)
• One-Step Diffusion Model for Image Motion-Deblurring [85.76149042561507]
  We propose a one-step diffusion model for deblurring (OSDD), a novel framework that reduces the denoising process to a single step.<n>To tackle fidelity loss in diffusion models, we introduce an enhanced variational autoencoder (eVAE), which improves structural restoration.<n>Our method achieves strong performance on both full and no-reference metrics.
  arXiv  Detail & Related papers  (2025-03-09T09:39:57Z)
• From Slow Bidirectional to Fast Autoregressive Video Diffusion Models [48.35054927704544]
  Current video diffusion models achieve impressive generation quality but struggle in interactive applications due to bidirectional attention dependencies.<n>We address this limitation by adapting a pretrained bidirectional diffusion transformer to an autoregressive transformer that generates frames on-the-fly.<n>Our model achieves a total score of 84.27 on the VBench-Long benchmark, surpassing all previous video generation models.
  arXiv  Detail & Related papers  (2024-12-10T18:59:50Z)
• Presto! Distilling Steps and Layers for Accelerating Music Generation [49.34961693154768]
  Presto! is an approach to inference acceleration for score-based diffusion transformers.<n>We develop a new score-based distribution matching distillation (DMD) method for the EDM-family of diffusion models.<n>To reduce the cost per step, we develop a simple, but powerful improvement to a recent layer distillation method.
  arXiv  Detail & Related papers  (2024-10-07T16:24:18Z)
• Adaptive Training Meets Progressive Scaling: Elevating Efficiency in Diffusion Models [52.1809084559048]
  We propose a novel two-stage divide-and-conquer training strategy termed TDC Training.<n>It groups timesteps based on task similarity and difficulty, assigning highly customized denoising models to each group, thereby enhancing the performance of diffusion models.<n>While two-stage training avoids the need to train each model separately, the total training cost is even lower than training a single unified denoising model.
  arXiv  Detail & Related papers  (2023-12-20T03:32:58Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.