Diffusion Model Inference Acceleration

Follow this topic Ask about this topic

Diffusion models repeatedly denoise a sample, so inference acceleration studies how to reduce the cost of those steps for image, video, language, and action generation. Methods reduce the number of evaluations with faster solvers, timestep schedules, and few-step distillation, or reduce work within each step using caching, sparse attention, token pruning, compression, and lower-precision computation. Systems work also targets memory movement, GPU parallelism, and serving, while adaptive methods allocate computation according to uncertainty, content, or task difficulty. The central challenge is preserving fidelity, diversity, and task accuracy as skipped or approximated computations can accumulate errors; gains therefore depend on the model, workload, hardware, and end-to-end evaluation.

Latest papers

509 papers73 papers in the last four weeks, up 66% on the four weeks before. 0.4% of all new papers.

Apr 2, 2026

Dependency-Guided Parallel Decoding in Discrete Diffusion Language Models

Discrete diffusion language models (dLLMs) accelerate text generation by unmasking multiple tokens in parallel. However, parallel decoding introduces a distributional mismatch: it approximates the joint conditional using a fully factorized product of per-token marginals, which degrades output quality when selected tokens are strongly dependent. We propose DEMASK (DEpendency-guided unMASKing), a lightweight dependency predictor that attaches to the final hidden states of a dLLM. In a single forward pass, it estimates pairwise conditional influences between masked positions. Using these predictions, a greedy selection algorithm identifies positions with bounded cumulative dependency for simultaneous unmasking. Under a sub-additivity assumption, we prove this bounds the total variation distance between our parallel sampling and the model's joint. Empirically, DEMASK achieves 1.7-2.2

\times

speedup on Dream-7B while matching or improving accuracy compared to confidence-based and KL-based baselines.

Liran Ringel, Ameen Ali, Yaniv Romano

Department of Computer Science, Technion – Israel Institute of Technology · Department of Electrical and Computer Engineering, Technion – Israel Institute of Technology · Blavatnik School of Computer Science and AI, Tel Aviv, Israel

Diffusion Language Model Decoding Diffusion Language Models Parallel Decoding Diffusion Model Inference Acceleration

Apr 2, 2026

ZEUS: Accelerating Diffusion Models with Only Second-Order Predictor

Denoising generative models deliver high-fidelity generation but remain bottlenecked by inference latency due to the many iterative denoiser calls required during sampling. Training-free acceleration methods reduce latency by either sparsifying the model architecture or shortening the sampling trajectory. Current training-free acceleration methods are more complex than necessary: higher-order predictors amplify error under aggressive speedups, and architectural modifications hinder deployment. Beyond 2x acceleration, step skipping creates structural scarcity -- at most one fresh evaluation per local window -- leaving the computed output and its backward difference as the only causally grounded information. Based on this, we propose ZEUS, an acceleration method that predicts reduced denoiser evaluations using a second-order predictor, and stabilizes aggressive consecutive skipping with an interleaved scheme that avoids back-to-back extrapolations. ZEUS adds essentially zero overhead, no feature caches, and no architectural modifications, and it is compatible with different backbones, prediction objectives, and solver choices. Across image and video generation, ZEUS consistently improves the speed-fidelity performance over recent training-free baselines, achieving up to 3.2x end-to-end speedup while maintaining perceptual quality. Our code is available at: https://github.com/Ting-Justin-Jiang/ZEUS.

Yixiao Wang, Ting Jiang, Zishan Shao +6

Department of Computer Science, Duke University, Durham, NC, USA · Department of Statistical Science, Duke University, Durham, NC, USA · Department of Electrical and Computer Engineering, Duke University, Durham, NC, USA

Diffusion Model Inference Acceleration Diffusion-Based Image Generation

Mar 30, 2026

EdgeDiT: Hardware-Aware Diffusion Transformers for Efficient On-Device Image Generation

Diffusion Transformers (DiT) have established a new state-of-the-art in high-fidelity image synthesis; however, their massive computational complexity and memory requirements hinder local deployment on resource-constrained edge devices. In this paper, we introduce EdgeDiT, a family of hardware-efficient generative transformers specifically engineered for mobile Neural Processing Units (NPUs), such as the Qualcomm Hexagon and Apple Neural Engine (ANE). By leveraging a hardware-aware optimization framework, we systematically identify and prune structural redundancies within the DiT backbone that are particularly taxing for mobile data-flows. Our approach yields a series of lightweight models that achieve a 20-30% reduction in parameters, a 36-46% decrease in FLOPs, and a 1.65-fold reduction in on-device latency without sacrificing the scaling advantages or the expressive capacity of the original transformer architecture. Extensive benchmarking demonstrates that EdgeDiT offers a superior Pareto-optimal trade-off between Frechet Inception Distance (FID) and inference latency compared to both optimized mobile U-Nets and vanilla DiT variants. By enabling responsive, private, and offline generative AI directly on-device, EdgeDiT provides a scalable blueprint for transitioning large-scale foundation models from high-end GPUs to the palm of the user.

Sravanth Kodavanti, Manjunath Arveti, Sowmya Vajrala +2

Samsung Research Institute Bangalore, India

Diffusion Transformer Diffusion-Based Image Generation Diffusion Model Inference Acceleration On-Device Inference Neural Network Optimization

Mar 30, 2026

Drift-AR: Single-Step Visual Autoregressive Generation via Anti-Symmetric Drifting

Autoregressive (AR)-Diffusion hybrid paradigms combine AR's structured semantic modeling with diffusion's high-fidelity synthesis, yet suffer from a dual speed bottleneck: the sequential AR stage and the iterative multi-step denoising of the diffusion vision decode stage. Existing methods address each in isolation without a unified principle design. We observe that the per-position prediction entropy of continuous-space AR models naturally encodes spatially varying generation uncertainty, which simultaneously governing draft prediction quality in the AR stage and reflecting the corrective effort required by vision decoding stage, which is not fully explored before. Since entropy is inherently tied to both bottlenecks, it serves as a natural unifying signal for joint acceleration. In this work, we propose Drift-AR, which leverages entropy signal to accelerate both stages: 1) for AR acceleration, we introduce Entropy-Informed Speculative Decoding that align draft-target entropy distributions via a causal-normalized entropy loss, resolving the entropy mismatch that causes excessive draft rejection; 2) for visual decoder acceleration, we reinterpret entropy as the physical variance of the initial state for an anti-symmetric drifting field -- high-entropy positions activate stronger drift toward the data manifold while low-entropy positions yield vanishing drift -- enabling single-step (1-NFE) decoding without iterative denoising or distillation. Moreover, both stages share the same entropy signal, which is computed once with no extra cost. Experiments on MAR, TransDiff, and NextStep-1 demonstrate 3.8-5.5

\times

speedup with genuine 1-NFE decoding, matching or surpassing original quality. Code will be available at https://github.com/aSleepyTree/Drift-AR.

Zhen Zou, Xiaoxiao Ma, Mingde Yao +3

University of Science and Technology of China · The Chinese University of Hong Kong, MMLAB · JD Explore Academy +1

Autoregressive Diffusion Visual Autoregressive Models One-Step Generative Modeling Speculative Decoding Diffusion Model Inference Acceleration

Mar 27, 2026

InstaVSR: Taming Diffusion for Efficient and Temporally Consistent Video Super-Resolution

Video super-resolution (VSR) seeks to reconstruct high-resolution frames from low-resolution inputs. While diffusion-based methods have substantially improved perceptual quality, extending them to video remains challenging for two reasons: strong generative priors can introduce temporal instability, and multi-frame diffusion pipelines are often too expensive for practical deployment. To address both challenges simultaneously, we propose InstaVSR, a lightweight diffusion framework for efficient video super-resolution. InstaVSR combines three ingredients: (1) a pruned one-step diffusion backbone that removes several costly components from conventional diffusion-based VSR pipelines, (2) recurrent training with flow-guided temporal regularization to improve frame-to-frame stability, and (3) dual-space adversarial learning in latent and pixel spaces to preserve perceptual quality after backbone simplification. On an NVIDIA RTX 4090, InstaVSR processes a 30-frame video at 2K

\times

2K resolution in under one minute with only 7 GB of memory usage, substantially reducing the computational cost compared to existing diffusion-based methods while maintaining favorable perceptual quality with significantly smoother temporal transitions.

Jintong Hu, Bin Chen, Zhenyu Hu +3

Insta360 Research · Peking University · Wuhan University

Video Super-Resolution Diffusion Models Diffusion Model Inference Acceleration

Mar 26, 2026

DRiffusion: Draft-and-Refine Process Parallelizes Diffusion Models with Ease

Diffusion models have achieved remarkable success in generating high-fidelity content but suffer from slow, iterative sampling, resulting in high latency that limits their use in interactive applications. We introduce DRiffusion, a parallel sampling framework that parallelizes diffusion inference through a draft-and-refine process. DRiffusion employs skip transitions to generate multiple draft states for future timesteps and computes their corresponding noises in parallel, which are then used in the standard denoising process to produce refined results. Theoretically, our method achieves an acceleration rate of

\tfrac{1}{n}

or

\tfrac{2}{n+1}

, depending on whether the conservative or aggressive mode is used, where

n

denotes the number of devices. Empirically, DRiffusion attains 1.4

\times

-3.7

\times

speedup across multiple diffusion models while incur minimal degradation in generation quality: on MS-COCO dataset, both FID and CLIP remain largely on par with those of the original model, while PickScore and HPSv2.1 show only minor average drops of 0.17 and 0.43, respectively. These results verify that DRiffusion delivers substantial acceleration and preserves perceptual quality.

Runsheng Bai, Chengyu Zhang, Yangdong Deng

CSAIL, MIT · IST, Nanjing University · School of Software, Tsinghua University; FuturististAI Lab, Shanghai Tsinghua International Innovation Center

Diffusion Model Sampling Diffusion Model Inference Acceleration Diffusion Sampling Diffusion-Based Image Generation Diffusion-Based Denoising

Mar 26, 2026

S2D2: Fast Decoding for Diffusion LLMs via Training-Free Self-Speculation

Block-diffusion language models offer a promising path toward faster-than-autoregressive generation by combining block-wise autoregressive decoding with within-block parallel denoising. However, in the few-step regime needed for practical acceleration, standard confidence-thresholded decoding is often brittle: aggressive thresholds hurt quality, while conservative thresholds require unnecessary denoising steps. Existing approaches that address this issue either require additional training or incur extra test-time compute. We present S2D2, a training-free self-speculative decoding framework for block-diffusion language models. Our key observation is that a block-diffusion model becomes autoregressive when the block size is reduced to one, allowing the same pretrained model to act as both drafter and verifier. S2D2 inserts a speculative verification step into standard block-diffusion decoding and uses lightweight routing policies to decide when verification is worth its cost. This yields a hybrid decoding trajectory in which diffusion proposes tokens in parallel, while the autoregressive mode acts as a local sequence-level critic. Across three mainstream block-diffusion families, S2D2 consistently improves the accuracy-speed tradeoff over strong confidence-thresholding baselines. On SDAR, we observe up to

4.7\times

speedup over autoregressive decoding, and up to

1.57\times

over a tuned dynamic decoding baseline while improving accuracy by up to

4.5

points. On LLaDA2.1-Mini, S2D2 remains complementary to built-in self-correction, including a conservative setting where it is

4.4\times

faster than the static baseline with slightly higher accuracy.

Ligong Han, Hao Wang, Han Gao +2

Red Hat AI Innovation · MIT-IBM Watson AI Lab · Iowa State University +1

Speculative Decoding for Diffusion Language Models Diffusion Language Model Decoding Diffusion Model Inference Acceleration Blockwise Diffusion Language Models Speculative Decoding

Mar 25, 2026

FODMP: Fast One-Step Diffusion of Movement Primitives Generation for Time-Dependent Robot Actions

Diffusion models are increasingly used for robot learning, but current designs face a clear trade-off. Action-chunking diffusion policies like ManiCM are fast to run, yet they only predict short segments of motion. This makes them reactive, but unable to capture time-dependent motion primitives, such as following a spring-damper-like behavior with built-in dynamic profiles of acceleration and deceleration. Recently, Movement Primitive Diffusion (MPD) partially addresses this limitation by parameterizing full trajectories using Probabilistic Dynamic Movement Primitives (ProDMPs), thereby enabling the generation of temporally structured motions. Nevertheless, MPD integrates the motion decoder directly into a multi-step diffusion process, resulting in prohibitively high inference latency that limits its applicability in real-time control settings. We propose FODMP (Fast One-step Diffusion of Movement Primitives), a new framework that distills diffusion models into the ProDMPs trajectory parameter space and generates motion using a single-step decoder. FODMP retains the temporal structure of movement primitives while eliminating the inference bottleneck through single-step consistency distillation. This enables robots to execute time-dependent primitives at high inference speed, suitable for closed-loop vision-based control. On standard manipulation benchmarks (MetaWorld, ManiSkill), FODMP runs up to 10 times faster than MPD and 7 times faster than action-chunking diffusion policies, while matching or exceeding their success rates. Beyond speed, by generating fast acceleration-deceleration motion primitives, FODMP allows the robot to intercept and securely catch a fast-flying ball, whereas action-chunking diffusion policy and MPD respond too slowly for real-time interception.

Xirui Shi, Arya Ebrahimi, Yi Hu +1

Department of Electrical and Computer Engineering, University of Alberta, Edmonton, Alberta, Canada. · Jun Jin is also with the Alberta Machine Intelligence Institute(Amii), Edmonton, Alberta, Canada.

Robot Motion Generation Diffusion Models for Motion Generation Diffusion Model Inference Acceleration

Mar 25, 2026

Calibri: Enhancing Diffusion Transformers via Parameter-Efficient Calibration

In this paper, we uncover the hidden potential of Diffusion Transformers (DiTs) to significantly enhance generative tasks. Through an in-depth analysis of the denoising process, we demonstrate that introducing a single learned scaling parameter can significantly improve the performance of DiT blocks. Building on this insight, we propose Calibri, a parameter-efficient approach that optimally calibrates DiT components to elevate generative quality. Calibri frames DiT calibration as a black-box reward optimization problem, which is efficiently solved using an evolutionary algorithm and modifies just ~100 parameters. Experimental results reveal that despite its lightweight design, Calibri consistently improves performance across various text-to-image models. Notably, Calibri also reduces the inference steps required for image generation, all while maintaining high-quality outputs.

Danil Tokhchukov, Aysel Mirzoeva, Andrey Kuznetsov +1

MSU · FusionBrain Lab, AXXX

Diffusion Transformer T2I Diffusion Models Diffusion Models Diffusion Model Inference Acceleration

Mar 25, 2026

Polynomial Speedup in Diffusion Models with the Multilevel Euler-Maruyama Method

We introduce the Multilevel Euler-Maruyama (ML-EM) method compute solutions of SDEs and ODEs using a range of approximators

f^1,\dots,f^k

to the drift

f

with increasing accuracy and computational cost, only requiring a few evaluations of the most accurate

f^k

and many evaluations of the less costly

f^1,\dots,f^{k-1}

. If the drift lies in the so-called Harder than Monte Carlo (HTMC) regime, i.e. it requires

ε^{-γ}

compute to be

ε

-approximated for some

γ>2

, then ML-EM

ε

-approximates the solution of the SDE with

ε^{-γ}

compute, improving over the traditional EM rate of

ε^{-γ-1}

. In other terms it allows us to solve the SDE at the same cost as a single evaluation of the drift. In the context of diffusion models, the different levels

f^{1},\dots,f^{k}

are obtained by training UNets of increasing sizes, and ML-EM allows us to perform sampling with the equivalent of a single evaluation of the largest UNet. Our numerical experiments confirm our theory: we obtain up to fourfold speedups for image generation on the CelebA dataset downscaled to 64x64, where we measure a

γ\approx2.5

. Given that this is a polynomial speedup, we expect even stronger speedups in practical applications which involve orders of magnitude larger networks.

Arthur Jacot

Diffusion Model Inference Acceleration Diffusion Sampling Stochastic Differential Equations Diffusion Models

Mar 25, 2026

Accelerating Diffusion-based Video Editing via Heterogeneous Caching: Beyond Full Computing at Sampled Denoising Timestep

Diffusion-based video editing has emerged as an important paradigm for high-quality and flexible content generation. However, despite their generality and strong modeling capacity, Diffusion Transformers (DiT) remain computationally expensive due to the iterative denoising process, posing challenges for practical deployment. Existing video diffusion acceleration methods primarily exploit denoising timestep-level feature reuse, which mitigates the redundancy in denoising process, but overlooks the architectural redundancy within the DiT that many attention operations over spatio-temporal tokens are redundantly executed, offering little to no incremental contribution to the model output. This work introduces HetCache, a training-free diffusion acceleration framework designed to exploit the inherent heterogeneity in diffusion-based masked video-to-video (MV2V) generation and editing. Instead of uniformly reuse or randomly sampling tokens, HetCache assesses the contextual relevance and interaction strength among various types of tokens in designated computing steps. Guided by spatial priors, it divides the spatial-temporal tokens in DiT model into context and generative tokens, and selectively caches the context tokens that exhibit the strongest correlation and most representative semantics with generative ones. This strategy reduces redundant attention operations while maintaining editing consistency and fidelity. Experiments show that HetCache achieves a noticeable acceleration, including a 2.67

\times

latency speedup and FLOPs reduction over commonly used foundation models, with negligible degradation in editing quality.

Tianyi Liu, Ye Lu, Linfeng Zhang +5

Nanyang Technological University · Shanghai Jiao Tong University · The Hong Kong Polytechnic University

Diffusion-Based Video Editing Diffusion Model Inference Acceleration Video Diffusion Models Video Editing Diffusion Model Caching Diffusion Transformer

Mar 23, 2026

Tiny Inference-Time Scaling with Latent Verifiers

Inference-time scaling has emerged as an effective way to improve generative models at test time by using a verifier to score and select candidate outputs. A common choice is to employ Multimodal Large Language Models (MLLMs) as verifiers, which can improve performance but introduce substantial inference-time cost. Indeed, diffusion pipelines operate in an autoencoder latent space to reduce computation, yet MLLM verifiers still require decoding candidates to pixel space and re-encoding them into the visual embedding space, leading to redundant and costly operations. In this work, we propose Verifier on Hidden States (VHS), a verifier that operates directly on intermediate hidden representations of Diffusion Transformer (DiT) single-step generators. VHS analyzes generator features without decoding to pixel space, thereby reducing the per-candidate verification cost while improving or matching the performance of MLLM-based competitors. We show that, under tiny inference budgets with only a small number of candidates per prompt, VHS enables more efficient inference-time scaling reducing joint generation-and-verification time by 63.3%, compute FLOPs by 51% and VRAM usage by 14.5% with respect to a standard MLLM verifier, achieving a +2.7% improvement on GenEval at the same inference-time budget.

Davide Bucciarelli, Evelyn Turri, Lorenzo Baraldi +3

University of Modena and Reggio Emilia, Italy · University of Pisa, Italy

Inference-Time Scaling Diffusion Models Diffusion Model Inference Acceleration

Mar 23, 2026

MinerU-Diffusion: Rethinking Document OCR as Inverse Rendering via Diffusion Decoding

Optical character recognition (OCR) has evolved from line-level transcription to structured document parsing, requiring models to recover long-form sequences containing layout, tables, and formulas. Despite recent advances in vision-language models, most existing systems rely on autoregressive decoding, which introduces sequential latency and amplifies error propagation in long documents. In this work, we revisit document OCR from an inverse rendering perspective, arguing that left-to-right causal generation is an artifact of serialization rather than an intrinsic property of the task. Motivated by this insight, we propose MinerU-Diffusion, a unified diffusion-based framework that replaces autoregressive sequential decoding with parallel diffusion denoising under visual conditioning. MinerU-Diffusion employs a block-wise diffusion decoder and an uncertainty-driven curriculum learning strategy to enable stable training and efficient long-sequence inference. Extensive experiments demonstrate that MinerU-Diffusion consistently improves robustness while achieving up to 3.2x faster decoding compared to autoregressive baselines. Evaluations on the proposed Semantic Shuffle benchmark further confirm its reduced dependence on linguistic priors and stronger visual OCR capability.

Hejun Dong, Junbo Niu, Bin Wang +3

Shanghai Artificial Intelligence Laboratory, OpenDataLab · Peking University

Optical Character Recognition Diffusion Language Model Decoding Blockwise Diffusion Language Models Diffusion Model Inference Acceleration

Mar 23, 2026

WorldCache: Content-Aware Caching for Accelerated Video World Models

Diffusion Transformers (DiTs) power high-fidelity video world models but remain computationally expensive due to sequential denoising and costly spatio-temporal attention. Training-free feature caching accelerates inference by reusing intermediate activations across denoising steps; however, existing methods largely rely on a Zero-Order Hold assumption i.e., reusing cached features as static snapshots when global drift is small. This often leads to ghosting artifacts, blur, and motion inconsistencies in dynamic scenes. We propose WorldCache, a Perception-Constrained Dynamical Caching framework that improves both when and how to reuse features. WorldCache introduces motion-adaptive thresholds, saliency-weighted drift estimation, optimal approximation via blending and warping, and phase-aware threshold scheduling across diffusion steps. Our cohesive approach enables adaptive, motion-consistent feature reuse without retraining. On Cosmos-Predict2.5-2B evaluated on PAI-Bench, WorldCache achieves \textbf{2.3

\times

} inference speedup while preserving 99.4% of baseline quality, substantially outperforming prior training-free caching approaches. Our code can be accessed on World-Cache.

Umair Nawaz, Ahmed Heakl, Ufaq Khan +3

Mohamed bin Zayed University of Artificial Intelligence, UAE · Linköping University, Sweden

Diffusion Model Caching Diffusion Model Inference Acceleration Video Diffusion Models Video World Models

Mar 23, 2026

Confidence-Based Decoding is Provably Efficient for Diffusion Language Models

Diffusion language models (DLMs) have emerged as a promising alternative to autoregressive (AR) models for language modeling, allowing flexible generation order and parallel generation of multiple tokens. However, this flexibility introduces a challenge absent in AR models: the decoding strategy -- which determines the order and number of tokens generated at each iteration -- critically affects sampling efficiency. Among decoding strategies explored in practice, confidence-based methods, which adaptively select which and how many tokens to unmask based on prediction confidence, have shown strong empirical performance. Despite this success, our theoretical understanding of confidence-based decoding remains limited. In this work, we develop the first theoretical analysis framework for confidence-based decoding in DLMs. We focus on an entropy sum-based strategy that continues unmasking tokens within each iteration until the cumulative entropy exceeds a threshold, and show that it achieves

\varepsilon

-accurate sampling in KL divergence with an expected number of iterations

\widetilde O(H(X_0)/\varepsilon)

, where

H(X_0)

denotes the entropy of the target data distribution. Notably, this strategy yields substantial sampling acceleration when the data distribution has low entropy relative to the sequence length, while automatically adapting to the intrinsic complexity of data without requiring prior knowledge or hyperparameter tuning. Overall, our results provide a theoretical foundation for confidence-based decoding and may inform the design of more efficient decoding strategies for DLMs.

Changxiao Cai, Gen Li

Department of Industrial and Operations Engineering, University of Michigan, Ann Arbor, USA · Department of Statistics and Data Science, The Chinese University of Hong Kong, Hong Kong

Diffusion Language Model Decoding Diffusion Model Inference Acceleration

Mar 23, 2026

DA-VAE: Plug-in Latent Compression for Diffusion via Detail Alignment

Reducing token count is crucial for efficient training and inference of latent diffusion models, especially at high resolution. A common strategy is to build high-compression image tokenizers with more channels per token. However, when trained only for reconstruction, high-dimensional latent spaces often lose meaningful structure, making diffusion training harder. Existing methods address this with extra objectives such as semantic alignment or selective dropout, but usually require costly diffusion retraining. Pretrained diffusion models, however, already exhibit a structured, lower-dimensional latent space; thus, a simpler idea is to expand the latent dimensionality while preserving this structure. We therefore propose Detail-Aligned VAE, which increases the compression ratio of a pretrained VAE with only lightweight adaptation of the pretrained diffusion backbone. DA-VAE uses an explicit latent layout: the first

C

channels come directly from the pretrained VAE at a base resolution, while an additional

D

channels encode higher-resolution details. A simple detail-alignment mechanism encourages the expanded latent space to retain the structure of the original one. With a warm-start fine-tuning strategy, our method enables

1024 \times 1024

image generation with Stable Diffusion 3.5 using only

32 \times 32

tokens,

4\times

fewer than the original model, within 5 H100-days. It further unlocks

2048 \times 2048

generation with SD3.5, achieving a

6\times

speedup while preserving image quality. We also validate the method and its design choices quantitatively on ImageNet.

Xin Cai, Zhiyuan You, Zhoutong Zhang +1

Multimedia Laboratory, The Chinese University of Hong Kong · Shanghai AI Laboratory · Adobe NextCam +1

Diffusion-Based Image Generation Diffusion Model Compression Diffusion Models Latent Diffusion Models Diffusion Model Inference Acceleration

Mar 23, 2026

Three Creates All: You Only Sample 3 Steps

Diffusion models deliver high-fidelity generation but remain slow at inference time due to many sequential network evaluations. We find that standard timestep conditioning becomes a key bottleneck for few-step sampling. Motivated by layer-dependent denoising dynamics, we propose Multi-layer Time Embedding Optimization (MTEO), which freeze the pretrained diffusion backbone and distill a small set of step-wise, layer-wise time embeddings from reference trajectories. MTEO is plug-and-play with existing ODE solvers, adds no inference-time overhead, and trains only a tiny fraction of parameters. Extensive experiments across diverse datasets and backbones show state-of-the-art performance in the few-step sampling and substantially narrow the gap between distillation-based and lightweight methods. Code will be available.

Yuren Cai, Guangyi Wang, Zongqing Li +3

School of Informatics, Xiamen University · School of Artificial Intelligence, Shenzhen Polytechnic University · Algorithm Research Department, Truesight

Few-Step Diffusion Sampling Diffusion Model Inference Acceleration Diffusion Model Distillation

Mar 22, 2026

Efficient Coarse-to-Fine Diffusion Models with Time Step Sequence Redistribution

Recently, diffusion models (DMs) have made significant strides in high-quality image generation. However, the multi-step denoising process often results in considerable computational overhead, impeding deployment on resource-constrained edge devices. Existing methods mitigate this issue by compressing models and adjusting the time step sequence. However, they overlook input redundancy and require lengthy search times. In this paper, we propose Coarse-to-Fine Diffusion Models with Time Step Sequence Redistribution. Recognizing indistinguishable early-stage generated images, we introduce Coarse-to-Fine Denoising (C2F) to reduce computation during coarse feature generation. Furthermore, we design Time Step Sequence Redistribution (TRD) for efficient sampling trajectory adjustment, requiring less than 10 minutes for search. Experimental results demonstrate that the proposed methods achieve near-lossless performance with an 80% to 90% reduction in computation on CIFAR10 and LSUN-Church.

Yu-Shan Tai, An-Yeu, Wu

Andy · Graduate Institute of Electrical Engineering, National Taiwan University, Taipei, Taiwan

Diffusion Model Sampling Diffusion Sampling Diffusion Model Inference Acceleration Diffusion Timestep Scheduling

Mar 20, 2026

Timestep-Aware Block Masking for Efficient Diffusion Model Inference

Diffusion Probabilistic Models (DPMs) have achieved great success in image generation but suffer from high inference latency due to their iterative denoising nature. Motivated by the evolving feature dynamics across the denoising trajectory, we propose a novel framework to optimize the computational graph of pre-trained DPMs on a per-timestep basis. By learning timestep-specific masks, our method dynamically determines which blocks to execute or bypass through feature reuse at each inference stage. Unlike global optimization methods that incur prohibitive memory costs via full-chain backpropagation, our method optimizes masks for each timestep independently, ensuring a memory-efficient training process. To guide this process, we introduce a timestep-aware loss scaling mechanism that prioritizes feature fidelity during sensitive denoising phases, complemented by a knowledge-guided mask rectification strategy to prune redundant spatial-temporal dependencies. Our approach is architecture-agnostic and demonstrates significant efficiency gains across a broad spectrum of models, including DDPM, LDM, DiT, and PixArt. Experimental results show that by treating the denoising process as a sequence of optimized computational paths, our method achieves a superior balance between sampling speed and generative quality. Our code will be released.

Haodong He, Yuan Gao, Weizhong Zhang +1

School of Computer Science, Wuhan University · School of Artificial Intelligence, Wuhan University · School of Data Science, Fudan University

Diffusion Model Acceleration Diffusion Model Inference Acceleration Diffusion-Based Image Generation

Mar 20, 2026

Accelerating Diffusion Decoders via Multi-Scale Sampling and One-Step Distillation

Image tokenization plays a central role in modern generative modeling by mapping visual inputs into compact representations that serve as an intermediate signal between pixels and generative models. Diffusion-based decoders have recently been adopted in image tokenization to reconstruct images from latent representations with high perceptual fidelity. In contrast to diffusion models used for downstream generation, these decoders are dedicated to faithful reconstruction rather than content generation. However, their iterative sampling process introduces significant latency, making them impractical for real-time or large-scale applications. In this work, we introduce a two-stage acceleration framework to address this inefficiency. First, we propose a multi-scale sampling strategy, where decoding begins at a coarse resolution and progressively refines the output by doubling the resolution at each stage, achieving a theoretical speedup of

\mathcal{O}(\log n)

compared to standard full-resolution sampling. Second, we distill the diffusion decoder at each scale into a single-step denoising model, enabling fast and high-quality reconstructions in a single forward pass per scale. Together, these techniques yield an order-of-magnitude reduction in decoding time with little degradation in output quality. Our approach provides a practical pathway toward efficient yet expressive image tokenizers. We hope it serves as a foundation for future work in efficient visual tokenization and downstream generation.

Chuhan Wang, Hao Chen

Carnegie Mellon University

Image Tokenization Diffusion Model Inference Acceleration Image Reconstruction Few-Step Diffusion Sampling Diffusion Model Distillation

Mar 19, 2026

Spectrally-Guided Diffusion Noise Schedules

Denoising diffusion models are widely used for high-quality image and video generation. Their performance depends on noise schedules, which define the distribution of noise levels applied during training and the sequence of noise levels traversed during sampling. Noise schedules are typically handcrafted and require manual tuning across different resolutions. In this work, we propose a principled way to design per-instance noise schedules for pixel diffusion, based on the image's spectral properties. By deriving theoretical bounds on the efficacy of minimum and maximum noise levels, we design ``tight'' noise schedules that eliminate redundant steps. During inference, we propose to conditionally sample such noise schedules. Experiments show that our noise schedules improve generative quality of single-stage pixel diffusion models, particularly in the low-step regime.

Carlos Esteves, Ameesh Makadia

Google Research.

Diffusion-Based Image Generation Diffusion Timestep Scheduling Diffusion-Based Denoising Pixel-Space Diffusion Models Diffusion Model Inference Acceleration

Mar 18, 2026

Attention-Based Sampler for Diffusion Language Models

Auto-regressive models (ARMs) have established a dominant paradigm in language modeling. However, their strictly sequential sampling paradigm imposes fundamental constraints on both inference efficiency and modeling flexibility. To address these limitations, diffusion-based large language models (dLLMs) have been proposed, offering the potential for parallel sampling and flexible language modeling. Despite these advantages, current dLLMs sampling strategies rely primarily on token level information, which fails to account for global sequence structure and often yields suboptimal results. In this paper, we study the sampling order selection problem from the perspective of log-likelihood maximization. We show that this problem is NP-hard and propose an optimal sampling-rank-based approximation that makes the objective computationally tractable. We further prove that the tractable objective is optimized by sampling tokens in descending order of their attention-matrix column sums. This finding provides a principled justification for attention-guided sampling and offers a theoretically grounded alternative to greedy search. We instantiate this theoretical insight in a new training-free sampling algorithm, termed Attn-Sampler, and further propose dynamic attention thresholding for practical acceleration. Extensive experiments across multiple benchmarks validate the effectiveness of our proposed method, demonstrating that it achieves superior generation quality while enhancing the sampling parallelism.

Yuyan Zhou, Kai Syun Hou, Weiyu Chen +1

Department of Computer Science and Engineering, The Hong Kong University of Science and Technology

Diffusion Sampling Diffusion Language Model Decoding Diffusion Language Models Diffusion Model Inference Acceleration

Mar 17, 2026

$D^3$ -RSMDE: 40 $\times$ Faster and High-Fidelity Remote Sensing Monocular Depth Estimation

Real-time, high-fidelity monocular depth estimation from remote sensing imagery is crucial for numerous applications, yet existing methods face a stark trade-off between accuracy and efficiency. Although using Vision Transformer (ViT) backbones for dense prediction is fast, they often exhibit poor perceptual quality. Conversely, diffusion models offer high fidelity but at a prohibitive computational cost. To overcome these limitations, we propose Depth Detail Diffusion for Remote Sensing Monocular Depth Estimation (

D^3

-RSMDE), an efficient framework designed to achieve an optimal balance between speed and quality. Our framework first leverages a ViT-based module to rapidly generate a high-quality preliminary depth map construction, which serves as a structural prior, effectively replacing the time-consuming initial structure generation stage of diffusion models. Based on this prior, we propose a Progressive Linear Blending Refinement (PLBR) strategy, which uses a lightweight U-Net to refine the details in only a few iterations. The entire refinement step operates efficiently in a compact latent space supported by a Variational Autoencoder (VAE). Extensive experiments demonstrate that

D^3

-RSMDE achieves a notable 11.85% reduction in the Learned Perceptual Image Patch Similarity (LPIPS) perceptual metric over leading models like Marigold, while also achieving over a 40x speedup in inference and maintaining VRAM usage comparable to lightweight ViT models.

Ruizhi Wang, Weihan Li, Zunlei Feng +5

School of Software Technology, Zhejiang University · State Key Laboratory of Blockchain and Data Security, Zhejiang University · Hangzhou High-Tech Zone (Binjiang) Institute of Blockchain and Data Security +2

Monocular Depth Estimation Diffusion Model Inference Acceleration Latent Diffusion Models

Mar 16, 2026

Flash-Unified: A Training-Free and Task-Aware Acceleration Framework for Native Unified Models

Native unified multimodal models, which integrate both generative and understanding capabilities, face substantial computational overhead that hinders their real-world deployment. Existing acceleration techniques typically employ a static, monolithic strategy, ignoring the fundamental divergence in computational profiles between iterative generation tasks (e.g., image generation) and single-pass understanding tasks (e.g., VQA). In this work, we present the first systematic analysis of unified models, revealing pronounced parameter specialization, where distinct neuron sets are critical for each task. This implies that, at the parameter level, unified models have implicitly internalized separate inference pathways for generation and understanding within a single architecture. Based on these insights, we introduce a training-free and task-aware acceleration framework, FlashU, that tailors optimization to each task's demands. Across both tasks, we introduce Task-Specific Network Pruning and Dynamic Layer Skipping, aiming to eliminate inter-layer and task-specific redundancy. For visual generation, we implement a time-varying control signal for the guidance scale and a temporal approximation for the diffusion head via Diffusion Head Cache. For multimodal understanding, building upon the pruned model, we introduce Dynamic Token Pruning via a V-Norm Proxy to exploit the spatial redundancy of visual inputs. Extensive experiments on Show-o2 demonstrate that FlashU achieves 1.78

\times

to 2.01

\times

inference acceleration across both understanding and generation tasks while maintaining SOTA performance, outperforming competing unified models and validating our task-aware acceleration paradigm. Our code is publicly available at https://github.com/Rirayh/FlashU.

Junlong Ke, Zichen Wen, Boxue Yang +6

Tsinghua University · Shanghai Jiao Tong University · Shanghai AI Laboratory +3

Efficient Multimodal Inference Unified Multimodal Models Diffusion Model Inference Acceleration Neural Network Pruning

Mar 16, 2026

Chain-of-Trajectories: Unlocking the Intrinsic Generative Optimality of Diffusion Models via Graph-Theoretic Planning

Diffusion models operate in a reflexive System 1 mode, constrained by a fixed, content-agnostic sampling schedule. This rigidity arises from the curse of state dimensionality, where the combinatorial explosion of possible states in the high-dimensional noise manifold renders explicit trajectory planning intractable and leads to systematic computational misallocation. To address this, we introduce Chain-of-Trajectories (CoTj), a train-free framework enabling System 2 deliberative planning. Central to CoTj is Diffusion DNA, a low-dimensional signature that quantifies per-stage denoising difficulty and serves as a proxy for the high-dimensional state space, allowing us to reformulate sampling as graph planning on a directed acyclic graph. Through a Predict-Plan-Execute paradigm, CoTj dynamically allocates computational effort to the most challenging generative phases. Experiments across multiple generative models demonstrate that CoTj discovers context-aware trajectories, improving output quality and stability while reducing redundant computation. This work establishes a new foundation for resource-aware, planning-based diffusion modeling. The code is available at https://github.com/UnicomAI/CoTj.

Ping Chen, Xiang Liu, Xingpeng Zhang +7

Data Science & Artificial Intelligence Research Institute, China Unicom · Unicom Data Intelligence, China Unicom · School of Computer Science and Software Engineering, Southwest Petroleum University +2

Diffusion Sampling Diffusion-Based Planning Diffusion Timestep Scheduling Diffusion Model Inference Acceleration

Mar 15, 2026

LatSearch: Latent Reward-Guided Search for Faster Inference-Time Scaling in Video Diffusion

The recent success of inference-time scaling in large language models has inspired similar explorations in video diffusion. In particular, motivated by the existence of "golden noise" that enhances video quality, prior work has attempted to improve inference by optimising or searching for better initial noise. However, these approaches have notable limitations: they either rely on priors imposed at the beginning of noise sampling or on rewards evaluated only on the denoised and decoded videos. This leads to error accumulation, delayed and sparse reward signals, and prohibitive computational cost, which prevents the use of stronger search algorithms. Crucially, stronger search algorithms are precisely what could unlock substantial gains in controllability, sample efficiency and generation quality for video diffusion, provided their computational cost can be reduced. To fill in this gap, we enable efficient inference-time scaling for video diffusion through latent reward guidance, which provides intermediate, informative and efficient feedback along the denoising trajectory. We introduce a latent reward model that scores partially denoised latents at arbitrary timesteps with respect to visual quality, motion quality, and text alignment. Building on this model, we propose LatSearch, a novel inference-time search mechanism that performs Reward-Guided Resampling and Pruning (RGRP). In the resampling stage, candidates are sampled according to reward-normalised probabilities to reduce over-reliance on the reward model. In the pruning stage, applied at the final scheduled step, only the candidate with the highest cumulative reward is retained, improving both quality and efficiency. We evaluate LatSearch on the VBench-2.0 benchmark and demonstrate that it consistently improves video generation across multiple evaluation dimensions compared to the baseline Wan2.1 model.

Zengqun Zhao, Ziquan Liu, Yu Cao +5

Queen Mary University of London · Huawei R&D UK · Imperial College London

Video Diffusion Models Inference-Time Scaling Diffusion Model Inference Acceleration

Mar 15, 2026

Early Failure Detection and Intervention in Video Diffusion Models

Text-to-video (T2V) diffusion models have rapidly advanced, yet generations still occasionally fail in practice, such as low text-video alignment or low perceptual quality. Since diffusion sampling is non-deterministic, it is difficult to know during inference whether a generation will succeed or fail, incurring high computational cost due to trial-and-error regeneration. To address this, we propose an early failure detection and diagnostic intervention pipeline for latent T2V diffusion models. For detection, we design a Real-time Inspection (RI) module that converts latents into intermediate video previews, enabling the use of established text-video alignment scorers for inspection in the RGB space. The RI module completes the conversion and inspection process in just 39.2ms. This is highly efficient considering that CogVideoX-5B requires 4.3s per denoising step when generating a 480p, 49-frame video on an NVIDIA A100 GPU. Subsequently, we trigger a hierarchical and early-exit intervention pipeline only when failure is predicted. Experiments on CogVideoX-5B and Wan2.1-1.3B demonstrate consistency gains on VBench with up to 2.64 times less time overhead compared to post-hoc regeneration. Our method also generalizes to a higher-capacity setting, remaining effective on Wan2.1-14B with 720p resolution and 81-frame generation. Furthermore, our pipeline is plug-and-play and orthogonal to existing techniques, showing seamless compatibility with prompt refinement and sampling guidance methods. We also provide evidence that failure signals emerge early in the denoising process and are detectable within intermediate video previews using standard vision-language evaluators.

Kwon Byung-Ki, Sohwi Lim, Nam Hyeon-Woo +2

Grad. School of Artificial Intelligence, POSTECH · School of Electrical Engineering, KAIST · Dept. of Electrical Engineering, POSTECH +1

Video Diffusion Models T2V Generation Diffusion Model Inference Acceleration

Mar 14, 2026

TMPDiff: Temporal Mixed-Precision for Diffusion Models

Diffusion models are the go-to method for Text-to-Image generation, but their iterative denoising processes has high inference latency. Quantization reduces compute time by using lower bitwidths, but applies a fixed precision across all denoising timesteps, leaving an entire optimization axis unexplored. We propose TMPDiff, a temporal mixed-precision framework for diffusion models that assigns different numeric precision to different denoising timesteps. We hypothesize that quantization errors accumulate additively across timesteps, which we then validate experimentally. Based on our observations, we develop an adaptive bisectioning-based algorithm, which assigns per-step precisions with linear evaluation complexity, reducing an otherwise exponential search problem. Across four state-of-the-art diffusion models and three datasets, TMPDiff consistently outperforms uniform-precision baselines at matched speedup, achieving 10 to 20% improvement in perceptual quality. On FLUX.1-dev, TMPDiff achieves 90% SSIM relative to the full-precision model at a speedup of 2.5x over 16-bit inference.

Basile Lewandowski, Simon Kurz, Aditya Shankar +3

University of Neuchâtel, Switzerland · TU Dortmund University, Germany · LAMARR Institute, Germany +2

Diffusion Model Quantization Diffusion Model Inference Acceleration Mixed-Precision Quantization

Mar 14, 2026

Amortizing Trajectory Diffusion with Keyed Drift Fields

Diffusion-based trajectory planners can synthesize rich, multimodal action sequences for offline reinforcement learning, but their iterative denoising incurs substantial inference-time cost, making closed-loop planning slow under tight compute budgets. We study the problem of achieving diffusion-like trajectory planning behavior with one-step inference, while retaining the ability to sample diverse candidate plans and condition on the current state in a receding-horizon control loop. Our key observation is that conditional trajectory generation fails under naïve distribution-matching objectives when the similarity measure used to align generated trajectories with the dataset is dominated by unconstrained future dimensions. In practice, this causes attraction toward average trajectories, collapses action diversity, and yields near-static behavior. Our key insight is that conditional generative planning requires a conditioning-aware notion of neighborhood: trajectory updates should be computed using distances in a compact key space that reflects the condition, while still applying updates in the full trajectory space. Building on this, we introduce Keyed Drifting Policies (KDP), a one-step trajectory generator trained with a drift-field objective that attracts generated trajectories toward condition-matched dataset windows and repels them from nearby generated samples, using a stop-gradient drifted target to amortize iterative refinement into training. At inference, the resulting policy produces a full trajectory window in a single forward pass. Across standard RL benchmarks and real-time hardware deployments, KDP achieves strong performance with one-step inference and substantially lower planning latency than diffusion sampling. Project website, code and videos: https://keyed-drifting.github.io/

Gokul Puthumanaillam, Melkior Ornik

University of Illinois Urbana-Champaign

Trajectory Generation Diffusion-Based Planning Diffusion Model Inference Acceleration Generative Modeling Offline RL

Mar 13, 2026

DiT-IC: Aligned Diffusion Transformer for Efficient Image Compression

Diffusion-based image compression has recently shown outstanding perceptual fidelity, yet its practicality is hindered by prohibitive sampling overhead and high memory usage. Most existing diffusion codecs employ U-Net architectures, where hierarchical downsampling forces diffusion to operate in shallow latent spaces (typically with only 8x spatial downscaling), resulting in excessive computation. In contrast, conventional VAE-based codecs work in much deeper latent domains (16x - 64x downscaled), motivating a key question: Can diffusion operate effectively in such compact latent spaces without compromising reconstruction quality? To address this, we introduce DiT-IC, an Aligned Diffusion Transformer for Image Compression, which replaces the U-Net with a Diffusion Transformer capable of performing diffusion in latent space entirely at 32x downscaled resolution. DiT-IC adapts a pretrained text-to-image multi-step DiT into a single-step reconstruction model through three key alignment mechanisms: (1) a variance-guided reconstruction flow that adapts denoising strength to latent uncertainty for efficient reconstruction; (2) a self-distillation alignment that enforces consistency with encoder-defined latent geometry to enable one-step diffusion; and (3) a latent-conditioned guidance that replaces text prompts with semantically aligned latent conditions, enabling text-free inference. With these designs, DiT-IC achieves state-of-the-art perceptual quality while offering up to 30x faster decoding and drastically lower memory usage than existing diffusion-based codecs. Remarkably, it can reconstruct 2048x2048 images on a 16 GB laptop GPU.

Junqi Shi, Ming Lu, Xingchen Li +3

School of Electronic Science and Engineering, Nanjing University, Nanjing 210023, China

Learned Image Compression Generative Image Compression Diffusion Model Inference Acceleration Diffusion Transformer