Just-in-Time: Training-Free Spatial Acceleration for Diffusion Transformers

• URL: http://arxiv.org/abs/2603.10744v2
• Date: Wed, 18 Mar 2026 13:10:33 GMT
• Title: Just-in-Time: Training-Free Spatial Acceleration for Diffusion Transformers
• Abstract summary: Diffusion Transformers have established a new state-of-the-art in image synthesis, but the high computational cost hampers their practical deployment.<n>We introduce Just-in-Time (JiT), a novel training-free framework that addresses this challenge by acceleration in the spatial domain.<n>JiT formulates a spatially approximated generative ordinary differential equation (ODE) that drives the full latent state evolution.
• Score: 15.69483083441184
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Diffusion Transformers have established a new state-of-the-art in image synthesis, but the high computational cost of iterative sampling severely hampers their practical deployment. While existing acceleration methods often focus on the temporal domain, they overlook the substantial spatial redundancy inherent in the generative process, where global structures emerge long before fine-grained details are formed. The uniform computational treatment of all spatial regions represents a critical inefficiency. In this paper, we introduce Just-in-Time (JiT), a novel training-free framework that addresses this challenge by acceleration in the spatial domain. JiT formulates a spatially approximated generative ordinary differential equation (ODE) that drives the full latent state evolution based on computations from a dynamically selected, sparse subset of anchor tokens. To ensure seamless transitions as new tokens are incorporated to expand the dimensions of the latent state, we propose a deterministic micro-flow, a simple and effective finite-time ODE that maintains both structural coherence and statistical correctness. Extensive experiments on the state-of-the-art FLUX.1-dev model demonstrate that JiT achieves up to a 7x speedup with nearly lossless performance, significantly outperforming existing acceleration methods and establishing a new and superior trade-off between inference speed and generation fidelity.

Related papers

• Learning Spatio-Temporal Foundation Models from Pure Synthetic Data [85.23980441738811]
  NeoST is a scalable synthetic-training corpus to real-world bias, a latent-space reasoning architecture that generates and iteratively refines multiple future trajectories, and latent-space objectives that emphasize structural dynamics and enable inference-time correction under distribution shifts.<n>Extensive experiments across diverse real-world benchmarks show that NeoST consistently outperforms existing STFMs in diverse real-worldtemporal systems.
  arXiv  Detail & Related papers  (2026-06-27T02:56:46Z)
• Drifting Models for Surrogate Flow Modeling [0.4925906256430175]
  We introduce a conditional architecture that performs drifting in a learned VAE latent space and uses label-aware masking to align generated samples with their boundary conditions.<n>Our model matches iterative diffusion in accuracy and flow consistency while running two orders of magnitude faster.
  arXiv  Detail & Related papers  (2026-06-05T17:35:09Z)
• Towards Physically Consistent 4D Scene Reconstruction for Closed-loop Autonomous Driving Simulation [19.055975606481635]
  High-fidelity street scene reconstruction is pivotal for end-to-end autonomous driving simulation.<n>Existing 3DGS methods and their 4D extensions fail to simultaneously achieve both.<n>We propose Orthogonal Projected Gradient (OPG) to restore spatial identifiability.
  arXiv  Detail & Related papers  (2026-05-20T11:09:53Z)
• Physics-Informed Neural PDE Solvers via Spatio-Temporal MeanFlow [28.327009081337746]
  We introduce S-Temporal MeanFlow, which functions as a novel PDE solver learning the finite-interval evolution of physical states.<n>We extend the original MeanFlow constraint from the temporal to thetemporal domain, coupling time evolution with spatial consistency.<n>This yields a unified framework naturally accommodating both time-dependent and stationary PDEs.
  arXiv  Detail & Related papers  (2026-05-09T12:29:10Z)
• Parallel Complex Diffusion for Scalable Time Series Generation [50.01609741902786]
  PaCoDi is a spectral-native architecture that decouples generative modeling in the frequency domain.<n>We show that PaCoDi outperforms existing baselines in both generation quality and inference speed.
  arXiv  Detail & Related papers  (2026-02-10T14:31:53Z)
• OSDEnhancer: Taming Real-World Space-Time Video Super-Resolution with One-Step Diffusion [64.10689934231165]
  Video super-resolution models (DMs) have demonstrated exceptional success in video super-resolution (VSR)<n>Their potential for space-time video super-resolution (STVSR) necessitates recovering realistic visual content from low to high-resolution but also improving the frame rate with coherent dynamics.<n>We propose OSDEnhancer, a framework that represents the first method to initialize real-world STVSR through an efficient one-step diffusion process.<n> Experiments demonstrate that the proposed method achieves state-of-the-art performance while maintaining superior capability in real-world scenarios.
  arXiv  Detail & Related papers  (2026-01-28T06:59:55Z)
• SDiT: Semantic Region-Adaptive for Diffusion Transformers [4.7254170106792035]
  Diffusion Transformers (DiTs) achieve state-of-the-art performance in text-to-image synthesis but remain computationally expensive due to the iterative nature of denoising and the quadratic cost of global attention.<n>We propose SDiT, a Semantic Region-Adaptive Diffusion Transformer that allocates computation according to regional complexity.
  arXiv  Detail & Related papers  (2026-01-18T06:43:36Z)
• RainDiff: End-to-end Precipitation Nowcasting Via Token-wise Attention Diffusion [64.49056527678606]
  We propose a Token-wise Attention integrated into not only the U-Net diffusion model but also the radar-temporal encoder.<n>Unlike prior approaches, our method integrates attention into the architecture without incurring the high resource cost typical of pixel-space diffusion.<n>Our experiments and evaluations demonstrate that the proposed method significantly outperforms state-of-the-art approaches, robustness local fidelity, generalization, and superior in complex precipitation forecasting scenarios.
  arXiv  Detail & Related papers  (2025-10-16T17:59:13Z)
• A-FloPS: Accelerating Diffusion Sampling with Adaptive Flow Path Sampler [21.134678093577193]
  A-FloPS is a principled, training-free framework for flow-based generative models.<n>We show that A-FloPS consistently outperforms state-of-the-art training-free samplers in both sample quality and efficiency.<n>With as few as $5$ function evaluations, A-FloPS achieves substantially lower FID and generates sharper, more coherent images.
  arXiv  Detail & Related papers  (2025-08-22T13:28:16Z)
• DiffuSeq-v2: Bridging Discrete and Continuous Text Spaces for Accelerated Seq2Seq Diffusion Models [58.450152413700586]
  We introduce a soft absorbing state that facilitates the diffusion model in learning to reconstruct discrete mutations based on the underlying Gaussian space. We employ state-of-the-art ODE solvers within the continuous space to expedite the sampling process. Our proposed method effectively accelerates the training convergence by 4x and generates samples of similar quality 800x faster.
  arXiv  Detail & Related papers  (2023-10-09T15:29:10Z)
• A Universal Framework for Quantum Dissipation:Minimally Extended State Space and Exact Time-Local Dynamics [5.221249829454763]
  dynamics of open quantum systems is formulated in a minimally extended state space. Time-local evolution equation is created in a mixed Liouville-Fock space.
  arXiv  Detail & Related papers  (2023-07-31T15:57:10Z)
• Learning to Accelerate Partial Differential Equations via Latent Global Evolution [64.72624347511498]
  Latent Evolution of PDEs (LE-PDE) is a simple, fast and scalable method to accelerate the simulation and inverse optimization of PDEs. We introduce new learning objectives to effectively learn such latent dynamics to ensure long-term stability. We demonstrate up to 128x reduction in the dimensions to update, and up to 15x improvement in speed, while achieving competitive accuracy.
  arXiv  Detail & Related papers  (2022-06-15T17:31:24Z)
• Hessian-Free High-Resolution Nesterov Acceleration for Sampling [55.498092486970364]
  Nesterov's Accelerated Gradient (NAG) for optimization has better performance than its continuous time limit (noiseless kinetic Langevin) when a finite step-size is employed. This work explores the sampling counterpart of this phenonemon and proposes a diffusion process, whose discretizations can yield accelerated gradient-based MCMC methods.
  arXiv  Detail & Related papers  (2020-06-16T15:07:37Z)

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.