Not All Tokens Need 40 Steps: Heterogeneous Step Allocation in Diffusion Transformers for Efficient Video Generation

• URL: http://arxiv.org/abs/2605.06892v1
• Date: Thu, 07 May 2026 19:49:47 GMT
• Title: Not All Tokens Need 40 Steps: Heterogeneous Step Allocation in Diffusion Transformers for Efficient Video Generation
• Abstract summary: We introduce motion-free inference that assigns varying step budgets to different redundant tokens based on their velocity dynamics.<n>To resolve the resulting sequence-length mismatch, HSA introduces a KV-cache mechanism that allows active tokens to attend to the full sequence.<n>We evaluate HSA on the Wan-2 and LTX-2 models for both text-to-video (T2V) and image-to-video (2V) generation.
• Score: 53.176242285107485
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Diffusion Transformers (DiTs) have achieved state-of-the-art video generation quality, but they incur immense computational cost because standard inference applies the same number of denoising steps uniformly to every token in the sequence. It is well known that human vision ignores vast amounts of redundant motion. Why, then, do our densest models treat every spatiotemporal token with equal priority? In this paper, we introduce Heterogeneous Step Allocation (HSA), a training-free inference algorithm that assigns varying step budgets to different spatiotemporal tokens based on their velocity dynamics. To resolve the resulting sequence-length mismatch without sacrificing global context, HSA introduces a KV-cache synchronization mechanism that allows active tokens to attend to the full sequence while entirely bypassing inactive tokens. Furthermore, we derive a cached Euler update that advances the latent states of skipped tokens in a single operation without additional model evaluations. We evaluate HSA on the Wan-2 and LTX-2 models for both text-to-video (T2V) and image-to-video (I2V) generation. Our results demonstrate that HSA significantly outperforms previous state-of-the-art caching methods and the vanilla Flow Matching baseline, especially at aggressive acceleration regimes (e.g., 50% and 25% runtimes). Crucially, HSA achieves a superior quality-runtime Pareto frontier without the need for expensive offline profiling, robustly preserving structural integrity and generation quality even under tight computational budgets. Project page: https://ernestchu.github.io/hsa

Related papers

• Adaptive Tokenisation Via Temporal Redundancy Masking And Latent Inpainting [18.71433885255431]
  Adaptive video tokenisation seeks to dynamically allocate token budgets based on the underlying visual complexity of a sequence.<n>We show that the latent space of a frozen continuous video tokeniser inherently encodes temporal redundancy that can be exploited directly.<n>We introduce a parameter-free adaptive token allocation mechanism that applies a fixed threshold to per-position temporal-L1 differences.
  arXiv  Detail & Related papers  (2026-06-04T13:31:12Z)
• SAID: Accelerating Diffusion-Based Language Models via Scaffold-Aware Iterative Decoding [11.233923872369202]
  Diffusion large language models (DLLMs) enable non-autoregressive computation generation by iteratively denoising corrupted tokens with bidirectional context.<n>We propose SAID, a Scaffold-Aware Iterative Decoding framework that acceleratesDLLMs by reallocating across tokens.<n>Experiments on LLaDA-8B and LLaDA 1.5 across math, coding, and knowledge benchmarks show that SAID significantly accelerates inference with a maximum speedup of 9.1x while maintaining competitive performance.
  arXiv  Detail & Related papers  (2026-06-03T14:56:42Z)
• Accelerating Diffusion-based Video Editing via Heterogeneous Caching: Beyond Full Computing at Sampled Denoising Timestep [37.62908191585867]
  HetCache is a training-free diffusion acceleration framework for video-to-video (MV2V) generation and editing.<n>It reduces redundant attention operations while maintaining editing consistency and fidelity.<n> Experiments show that HetCache achieves a noticeable acceleration, including a 2.67$times$ latency speedup and FLOPs reduction over commonly used foundation models.
  arXiv  Detail & Related papers  (2026-03-25T12:53:31Z)
• EVATok: Adaptive Length Video Tokenization for Efficient Visual Autoregressive Generation [80.13014959623452]
  EVATok is a framework to produce $textbfE$fficient $textbfV$ideo $textbfA$daptive $textbfTok$enizers.<n>We show that EVATok achieves superior reconstruction and state-of-the-art class-to-video generation on UCF-101.
  arXiv  Detail & Related papers  (2026-03-12T17:59:59Z)
• AdaTSQ: Pushing the Pareto Frontier of Diffusion Transformers via Temporal-Sensitivity Quantization [22.45250803905198]
  Diffusion Transformers (DiTs) have emerged as the state-of-the-art backbone for high-fidelity image and video generation.<n>Post-training quantization (PTQ) has proven effective for large language models (LLMs)<n>We propose AdaTSQ, a novel PTQ framework that pushes the frontier of efficiency and quality by exploiting the temporal sensitivity of DiTs.
  arXiv  Detail & Related papers  (2026-02-10T15:23:18Z)
• OmniSAT: Compact Action Token, Faster Auto Regression [70.70037017501357]
  We introduce an Omni Swift Action Tokenizer, which learns a compact, transferable action representation.<n>The resulting discrete tokenization shortens the training sequence by 6.8$times$, and lowers the target entropy.
  arXiv  Detail & Related papers  (2025-10-08T03:55:24Z)
• Compute Only 16 Tokens in One Timestep: Accelerating Diffusion Transformers with Cluster-Driven Feature Caching [11.75972316736487]
  We introduce Cluster-Driven Feature Caching (ClusCa) to accelerate diffusion transformers.<n>ClusCa performs spatial clustering on tokens in each timestep, computes only one token in each cluster and propagates their information to all the other tokens.<n>Experiments on DiT, FLUX and HunyuanVideo demonstrate its effectiveness in both text-to-image and text-to-video generation.
  arXiv  Detail & Related papers  (2025-09-12T14:53:45Z)
• Sparse VideoGen2: Accelerate Video Generation with Sparse Attention via Semantic-Aware Permutation [84.00166854547241]
  Diffusion Transformers (DiTs) are essential for video generation but suffer from significant latency due to the quadratic complexity of attention.<n>We propose SVG2, a training-free framework that maximizes identification accuracy and computation minimizes waste.
  arXiv  Detail & Related papers  (2025-05-24T21:30:29Z)
• Prune Spatio-temporal Tokens by Semantic-aware Temporal Accumulation [89.88214896713846]
  STA score considers two critical factors: temporal redundancy and semantic importance. We apply the STA module to off-the-shelf video Transformers and Videowins. Results: Kinetics-400 and Something-Something V2 achieve 30% overshelf reduction with a negligible 0.2% accuracy drop.
  arXiv  Detail & Related papers  (2023-08-08T19:38:15Z)
• H$_2$O: Heavy-Hitter Oracle for Efficient Generative Inference of Large Language Models [110.06476624089679]
  We introduce a novel approach for implementing the KV cache which significantly reduces its memory footprint. Our approach is based on the observation that a small portion of tokens contributes most of the value when computing attention scores. We propose Heavy Hitter (H$$O), a KV cache eviction policy that dynamically retains a balance of recent and H$$ tokens.
  arXiv  Detail & Related papers  (2023-06-24T20:11:14Z)
• Paraformer: Fast and Accurate Parallel Transformer for Non-autoregressive End-to-End Speech Recognition [62.83832841523525]
  We propose a fast and accurate parallel transformer, termed Paraformer. It accurately predicts the number of output tokens and extract hidden variables. It can attain comparable performance to the state-of-the-art AR transformer, with more than 10x speedup.
  arXiv  Detail & Related papers  (2022-06-16T17:24:14Z)

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.