Pith. sign in

REVIEW 3 major objections 5 minor 1 cited by

SmoothCache: A Universal Inference Acceleration Technique for Diffusion Transformers

T0 review · 3 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read SmoothCache: a training-free calibration pass selects layers to cache during diffusion inference, yielding 8–71% speedups while preserving image, video, and audio quality.

desk verdict A simple, training-free DiT caching rule with a real multi-modal payoff; the quality claims are a bit too rosy and the calibration assumption is unvalidated, but the core idea is solid and testable. read the letter →

arxiv 2411.10510 v2 pith:ZEW7AX3D submitted 2024-11-15 cs.LG

Joseph Liu , Joshua Geddes , Ziyu Guo , Haomiao Jiang , Mahesh Kumar Nandwana This is my paper · ORCID
classification cs.LG
keywords diffusiontransformersinferenceaccelerationlayercachingtraining-freecross-timestepsimilaritycalibrationmultimodalgenerationresidualconnections
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper sets out to show that redundancy between neighboring denoising steps in Diffusion Transformers can be harvested by a universal, training-free caching rule. SmoothCache runs a single calibration inference pass on about ten samples, records per-layer L1 relative errors between outputs at nearby timesteps, and then during generation reuses cached layer outputs whenever the average error for that layer type sits below one threshold alpha. The authors argue the error curves are nearly sample-independent, so a schedule fixed at calibration time works for new prompts across image, video, and audio. If this holds, it would give practitioners a drop-in acceleration method for DiT models that needs no retraining, no data access, and no per-model hand-tuning.

What carries the argument

The load-bearing object is the layer-wise representation error curve $E(L_t, L_{t+k}) = \|L_t - L_{t+k}\|_1 / \|L_t\|_1$, computed on a small calibration set with caching disabled. SmoothCache averages this error over all layers of the same type and caches the whole group when the average is below $\alpha$, which prevents errors from cascading between layers of one type. The residual connection is what makes reuse safe: a cached block output is added to the unchanged residual stream, so the introduced error is the difference between two similar outputs rather than a full recomputation error.

What would settle it

Run SmoothCache on a held-out set of prompts with the calibrated alpha, but also recompute the skipped layers in the background so the true per-layer L1 relative error can be measured. If the true errors exceed alpha on many skipped steps, or if FID, VBench, or CLAP metrics degrade outside the paper's reported ranges at the recommended alpha, then the calibration-transfer assumption fails.

Watch

Extended reading notes

Core claim

Layer outputs of DiT blocks at adjacent diffusion timesteps are highly similar, and the L1 relative error between them is stable across different inputs. SmoothCache exploits this by caching the outputs of self-attention, cross-attention, and feed-forward layers that sit before residual connections, then reusing those outputs k timesteps later when the calibration error falls below a scalar threshold alpha. On DiT-XL-256x256, Open-Sora, and Stable Audio Open, the paper reports 8% to 71% end-to-end latency speedups while keeping FID, VBench, and CLAP metrics at or slightly better than the no-cache baseline, and it matches or outperforms modality-specific caching methods such as FORA and Learning-to-Cache at equal latency, without their training or architecture assumptions.

Load-bearing premise

The load-bearing premise is that the layer-error curves measured on ten calibration samples with caching switched off remain faithful estimates of the true per-layer errors during SmoothCache-enabled inference, and stay stable enough across new inputs that the same skip decisions are safe.

Editorial extensions

If this is right

  • Any DiT-based pipeline can adopt SmoothCache without retraining or access to training data; the method only needs one calibration pass and a scalar threshold.
  • The caching schedule is fixed at calibration time, so it is compatible with graph-compilation and other static inference optimizations.
  • A single $\alpha$ gives a fine-grained speed/quality dial, allowing practitioners to interpolate between no caching and aggressive caching.
  • At matched latency, SmoothCache reports better FID/sFID than the static FORA schedule on DiT-XL, and it extends to video and audio where FORA's fixed schedule fails.
  • On architecture–modality pairs with higher calibration error variance, such as Open-Sora, the available speedup is smaller, matching the paper's observed correlation between error variance and the width of the speed/quality front.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • An implicit consequence is that SmoothCache can be composed with step-reduction techniques: because it cuts per-step cost rather than the number of steps, combining it with fast ODE solvers or distillation should roughly multiply the speedups.
  • The same calibration-error principle might transfer to other iterative transformer architectures with residual connections and temporally correlated inputs, but that would require re-validating the stability of error curves in each new setting.
  • A practical extension suggested by the method is to calibrate $\alpha$ from a target latency or MACs budget by inverting the measured speedup curve, avoiding a search over quality metrics.
  • The paper's variance-correlation observation could be used as a cheap deployment diagnostic: measure the inter-sample spread of calibration curves and lower $\alpha$ when spread is high.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper proposes SmoothCache, a training-free, calibration-based caching scheme for Diffusion Transformer (DiT) inference. For a given DiT model, a small calibration set is used to compute average L1 relative errors between layer outputs at adjacent diffusion timesteps, and a single threshold α decides when a layer's output can be replaced by a cached output from an earlier timestep. The method is evaluated on DiT-XL for image generation, OpenSora for text-to-video, and Stable Audio Open for text-to-audio, reporting MACs and latency reductions of roughly 8% to 71% depending on the model and threshold. The paper claims this maintains or even improves generation quality, and compares favorably against FORA and Learning-to-Cache on image generation.

Significance. If the central claim is correct, SmoothCache would be a useful, architecture-agnostic inference acceleration technique that needs only one calibration pass and a single hyperparameter, with no retraining. The paper's strengths include evaluation across three modalities with standard metrics, comparison to concurrent caching methods, ablations over sampling steps and calibration set size, and a clear statement of the method's limitations. However, the abstract's 'maintaining or even improving generation quality' claim is not supported by the reported OpenSora and Stable Audio results, and the validity of the calibration error curves under cached inference is not directly verified. These issues are fixable, but they currently prevent the paper from fully establishing its main claims.

major comments (3)
  1. [Abstract; Section 3.2.1; Tables 2 and 3] The claim that SmoothCache maintains or improves generation quality across modalities is contradicted by the reported numbers. In Table 2, OpenSora VBench drops from 79.36 ± 0.19 to 78.76 ± 0.38 at α=0.02 and to 78.10 ± 0.51 at α=0.03, and the LPIPS/PSNR/SSIM values (e.g., PSNR 11.08 and 12.62) indicate large deviations from the non-cached outputs. In Table 3, Stable Audio Open FD OpenL3 increases on AudioCaps (81.7 → 84.5 → 89.6) and Song Describer (105.2 → 106.2 → 131.3), and KL PaSST and CLAP generally worsen at α=0.30. The 'maintaining or even improving' phrasing should be restricted to the settings where it actually holds, such as several DiT-XL configurations, or replaced by an explicit statement of the quality/speed tradeoff.
  2. [Section 2.2, Eq. (4); Section 4] The scheduling decision in Eq. (4) uses layer representation errors measured on a calibration pass with no caching, but during SmoothCache inference the inputs to downstream layers are themselves approximated by cached outputs from earlier timesteps. As the authors acknowledge in Section 4, 'the calibration loss is computed when no caching is performed, which may not fully model true approximation errors during SmoothCache-enabled inference.' This is load-bearing because every caching decision derives from these curves. The paper should provide direct evidence that the calibration curves remain faithful predictors of actual per-layer errors under caching, for example by measuring true per-layer errors during SmoothCache-enabled inference on a held-out set and comparing them to the calibration estimates, or by comparing the SmoothCache schedule against an oracle schedule built from actual cached-inference errors.
  3. [Section 2.2; Section 3.2; Tables 1-3] The selection of α, the only hyperparameter controlling the speed/quality operating point, is not described with a concrete protocol. The paper mentions a 'brief linear search' in Section 2.2, but the experimental section reports only the final α values chosen for each table, with no discussion of the search range, the criterion used to stop, or sensitivity to α. Because the reported tradeoffs are the central empirical contribution, the authors should specify how α was selected and report quality metrics across a range of α values, even for the modalities where only two operating points are shown.
minor comments (5)
  1. [Section 1.1] There are typographical issues in the related work section, including 'Initally' and the duplicated phrase 'such as such as'; these should be corrected.
  2. [Section 2.2, Eq. (4)] The notation in Eq. (4) is confusing: the left-hand side compares L_ij,t with L_ij,t+k, while the right-hand side averages over j terms using tilde quantities. Please clarify whether the approximation is the average over all layers of type i, and define N explicitly as the number of layers of that type.
  3. [Table 2] For OpenSora, LPIPS, PSNR, and SSIM are computed relative to non-cached videos, but no baseline row is provided (e.g., LPIPS=0, PSNR=∞, SSIM=1 for no cache). Reporting these values without the baseline makes the magnitude of the distortion hard to interpret.
  4. [Figure 2] All error curves are 'scaled to the same y-axis range,' which prevents visual comparison of absolute error magnitudes across components and models; please report the actual y-axis scales or include separate panels with raw scales.
  5. [Section 3.3] The ablation section states that 10 calibration samples are 'usually enough to reliably regenerate the same caching schedule,' but no quantitative evidence is given; reporting schedule similarity or quality metrics for 1, 5, 10, and 20 calibration samples would strengthen this claim.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: SmoothCache's caching schedule is an empirical calibration-and-threshold heuristic, and the reported speedups and quality trade-offs are measured on held-out test sets.

full rationale

The derivation chain is self-contained. SmoothCache measures per-layer L1 relative representation errors on a small calibration set without caching (Eq. 4), averages them over layers and samples, and compares them to a single hyperparameter alpha to decide which layer types are cache-eligible at each timestep. The claimed outputs, speedup and generation quality, are not computed from Eq. 4 but measured end-to-end on held-out evaluation sets (ImageNet FID/IS, VBench for OpenSora, CLAP/FD/KL for Stable Audio Open) against non-cached baselines and existing methods such as FORA and L2C. Alpha is a hyperparameter selected by linear search per sampling configuration, and multiple operating points are reported rather than a single fitted value being relabeled as a prediction. The paper explicitly acknowledges that calibration errors are computed when no caching is performed and may not fully model true errors during cached inference (Section 4), which is an honest limitation about miscalibration risk rather than a circular reduction. There are no load-bearing self-citations: prior caching works are cited only as related work and baselines, and the cross-timestep similarity observation is attributed to external sources [20,30]. No equation in the paper defines or derives its headline result from the same quantity it claims to predict. The miscalibration concern raised by the skeptic is a correctness risk, not circularity, and does not raise the circularity score under the stated rules.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The method introduces no new physical entities; it relies on standard diffusion assumptions plus an ad hoc calibration-error-to-cached-error equivalence that is self-acknowledged as fragile.

free parameters (3)
  • alpha (caching threshold) = 0.08, 0.18, 0.22 for DiT-XL; 0.02, 0.03 for OpenSora; 0.15, 0.30 for Stable Audio Open
    Single threshold on mean L1 relative error; hand-selected per model/solver/steps to realize the reported speed/quality tradeoff.
  • cache horizon k = 1-3 for DiT-XL and Stable Audio; up to 5 for OpenSora
    Chosen from observation that error grows too large beyond these gaps; a free choice per model.
  • calibration set size = 10 samples
    Claimed sufficient via ablation in Sec 3.3 and Supplementary; not systematically optimized.
assumptions (4)
  • domain assumption Layer outputs at adjacent diffusion timesteps are highly similar
    Cited in Sec 2.1 and observed in Fig 2; this is the basis for all caching; it is well established for diffusion models.
  • ad hoc to paper The average L1 relative error from 10 calibration samples predicts per-sample error curves and remains valid when layers are cached
    Sec 2.2 Eq. 4; explicitly acknowledged as a limitation in Sec 4 because computed without caching.
  • ad hoc to paper Grouping caching decisions by layer type prevents cascading errors
    Sec 2.2; the authors note this only partially resolves dependency issues between different layer types (Sec 4).
  • domain assumption Cached outputs can be injected before residual connections with negligible distortion
    Sec 2.2 and Fig 3; the paper states residual connections are 'essential for its caching effectiveness' (Sec 4).

how reviews work

0 comments
Cite this review

Pith. "Pith review of SmoothCache: A Universal Inference Acceleration Technique for Diffusion Transformers." pith.science (2026). https://pith.science/paper/ZEW7AX3D

@misc{pith2026241110510,
  author       = {Pith},
  title        = {Pith review of: SmoothCache: A Universal Inference Acceleration Technique for Diffusion Transformers},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZEW7AX3D}},
  note         = {Machine review of arXiv:2411.10510}
}
read the original abstract

Diffusion Transformers (DiT) have emerged as powerful generative models for various tasks, including image, video, and speech synthesis. However, their inference process remains computationally expensive due to the repeated evaluation of resource-intensive attention and feed-forward modules. To address this, we introduce SmoothCache, a model-agnostic inference acceleration technique for DiT architectures. SmoothCache leverages the observed high similarity between layer outputs across adjacent diffusion timesteps. By analyzing layer-wise representation errors from a small calibration set, SmoothCache adaptively caches and reuses key features during inference. Our experiments demonstrate that SmoothCache achieves 8% to 71% speed up while maintaining or even improving generation quality across diverse modalities. We showcase its effectiveness on DiT-XL for image generation, Open-Sora for text-to-video, and Stable Audio Open for text-to-audio, highlighting its potential to enable real-time applications and broaden the accessibility of powerful DiT models.

Figures

Figures reproduced from arXiv: 2411.10510 by the authors.

Figure 1
Figure 1. Accelerating Diffusion Transformer inference across multiple modalities with 50 DDIM Steps on DiT-XL-256x256, 100 DPM [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. L1 Relative Error Curves of different architecture components. Curves are plotted with 95% confidence intervals from 10 calibration samples from all components explored in this paper and scaled to the same y-axis range. Note that OpenSora has distinct spatial and temporal diffusion blocks. introducing a training-free caching approach, that makes no underlying assumption about the specific task or modality at hand, o… view at source ↗
Figure 3
Figure 3. Illustration of SmoothCache. When the layer representation loss obtained from the calibration pass is below some threshold α, the corresponding layer is cached and used in place of the same computation on a future timestep. The figure on the left shows how the layer representation error impacts whether certain layers are eligible for caching. The error of the attention (attn) layer is higher in earlier timesteps, so… view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4 [PITH_FULL_IMAGE:figures/full_fig_p004_4.png]
Figure 5
Figure 5. Figure 5: Layer Compute Composition of candidate models. These are computed from the MACs of the default model con￾figurations without SmoothCache applied. Note that in all candi￾date models, SmoothCache eligible layers comprise at least 90% of compute time. nential search space…
Figure 6
Figure 6. Figure 6: SmoothCache results on DiT-XL/2-256×256 for unconditional generation with 50 DDIM sampling steps on ImageNet-1k for thresholds 0.08 and 0.18, as well as for Static Caching. conditional prompts. For audio, we show the log-mel spec￾trogram in order to visualize the wavef…
Figure 7
Figure 7. Figure 7: SmoothCache Results on Stable Audio Open for threshold 0.15 and 0.3. Log-Mel Spectrograms are shown. ule given the same α. We however note that the confidence interval of the different error curves vary from modality to modality as seen in [PITH_FULL_IMAGE:figures/ful…
Figure 8
Figure 8. Figure 8: SmoothCache Results on OpenSora for threshold 0.03 for 2s 480p videos. We show the first, middle and last frame of each video. We use the following prompts, in order from top to bottom: (1) Chocolate sauce is poured slowly over a stack of fluffy pancakes. (2) an astron…
Figure 9
Figure 9. Figure 9: L1 Relative Error Curves of different architecture components for DiT-XL. Curves are plotted with 95% confidence intervals and scaled to the same y-axis range. 6. Ablations Addendum: Number of Calibra￾tion Samples As mentioned in Sec. 3.3, the calibration sample size c…

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Kaleido: Algorithm-Hardware Co-Design for Video Diffusion Transformers by Exploiting Latent Space Correlations

    cs.AR 2026-07 conditional novelty 6.0 of 10

    A channel-wise reuse algorithm plus a reconfigurable systolic accelerator skips redundant vDiT attention and MLP computation, achieving up to 5.9x speedup and 16x energy savings.

Reference graph

Works this paper leans on

42 extracted references · 20 canonical work pages · cited by 1 Pith paper

  1. [1]

    Ap- proximate caching for efficiently serving diffusion models

    Shubham Agarwal, Subrata Mitra, Sarthak Chakraborty, Srikrishna Karanam, Koyel Mukherjee, and Shiv Saini. Ap- proximate caching for efficiently serving diffusion models. arXiv preprint arXiv:2312.04429, 2023. 2

    arXiv 2023

  2. [2]

    Andrea Agostinelli, Timo I. Denk, Zal ´an Borsos, Jesse En- gel, Mauro Verzetti, Antoine Caillon, Qingqing Huang, Aren Jansen, Adam Roberts, Marco Tagliasacchi, Matt Sharifi, Neil Zeghidour, and Christian Frank. Musiclm: Generating music from text, 2023. 5

    work page 2023

  3. [3]

    Pixart-$\alpha$: Fast training of diffusion transformer for photorealistic text-to-image syn- thesis

    Junsong Chen, Jincheng YU, Chongjian GE, Lewei Yao, Enze Xie, Zhongdao Wang, James Kwok, Ping Luo, Huchuan Lu, and Zhenguo Li. Pixart-$\alpha$: Fast training of diffusion transformer for photorealistic text-to-image syn- thesis. In The Twelfth International Conference on Learning Representations, 2024. 1

    2024

  4. [4]

    δ-dit: A training-free acceleration method tailored for diffusion transformers

    Pengtao Chen, Mingzhu Shen, Peng Ye, Jianjian Cao, Chongjun Tu, Christos-Savvas Bouganis, Yiren Zhao, and Tao Chen. δ-dit: A training-free acceleration method tailored for diffusion transformers. arXiv preprint arXiv:2406.01125,

    arXiv

  5. [5]

    Stable audio open

    Zach Evans, Julian D Parker, CJ Carr, Zack Zukowski, Josiah Taylor, and Jordi Pons. Stable audio open. arXiv preprint arXiv:2407.14358, 2024. 1, 2, 5

    arXiv 2024

  6. [6]

    Structural pruning for diffusion models, 2023

    Gongfan Fang, Xinyin Ma, and Xinchao Wang. Structural pruning for diffusion models, 2023. 1, 2

    work page 2023

  7. [7]

    Ptqd: Accurate post-training quantization for diffusion models

    Yefei He, Luping Liu, Jing Liu, Weijia Wu, Hong Zhou, and Bohan Zhuang. Ptqd: Accurate post-training quantization for diffusion models. Advances in Neural Information Pro- cessing Systems, 36, 2024. 1, 2

    work page 2024

  8. [8]

    Denoising dif- fusion probabilistic models

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising dif- fusion probabilistic models. Advances in neural information processing systems, 33:6840–6851, 2020. 1, 2

    2020

Show all 42 references
  1. [9]

    On exact inversion of dpm-solvers,

    Seongmin Hong, Kyeonghyun Lee, Suh Yoon Jeon, Hyewon Bae, and Se Young Chun. On exact inversion of dpm-solvers,

  2. [10]

    VBench: Com- prehensive benchmark suite for video generative models

    Ziqi Huang, Yinan He, Jiashuo Yu, Fan Zhang, Chenyang Si, Yuming Jiang, Yuanhan Zhang, Tianxing Wu, Qingyang Jin, Nattapol Chanpaisit, Yaohui Wang, Xinyuan Chen, Limin Wang, Dahua Lin, Yu Qiao, and Ziwei Liu. VBench: Com- prehensive benchmark suite for video generative models....

    2024

  3. [11]

    Audiocaps: Generating captions for audios in the wild

    Chris Dongjoo Kim, Byeongchang Kim, Hyunmin Lee, and Gunhee Kim. Audiocaps: Generating captions for audios in the wild. In NAACL-HLT, 2019. 5

    2019

  4. [12]

    Open-sora-plan, 2024

    PKU-Yuan Lab and Tuzhan AI etc. Open-sora-plan, 2024. 1, 2

    2024

  5. [13]

    V oicebox: Text-guided multilingual universal speech generation at scale

    Matthew Le, Apoorv Vyas, Bowen Shi, Brian Karrer, Leda Sari, Rashel Moritz, Mary Williamson, Vimal Manohar, Yossi Adi, Jay Mahadeokar, et al. V oicebox: Text-guided multilingual universal speech generation at scale. Advances in neural information processing systems , 36, 2024. 1, 2

    2024

  6. [14]

    Snap- fusion: Text-to-image diffusion model on mobile devices within two seconds

    Yanyu Li, Huan Wang, Qing Jin, Ju Hu, Pavlo Chemerys, Yun Fu, Yanzhi Wang, Sergey Tulyakov, and Jian Ren. Snap- fusion: Text-to-image diffusion model on mobile devices within two seconds. Advances in Neural Information Pro- cessing Systems, 36, 2024. 1, 2

    2024

  7. [15]

    Vit- tts: visual text-to-speech with scalable diffusion transformer

    Huadai Liu, Rongjie Huang, Xuan Lin, Wenqiang Xu, Mao- zong Zheng, Hong Chen, Jinzheng He, and Zhou Zhao. Vit- tts: visual text-to-speech with scalable diffusion transformer. arXiv preprint arXiv:2305.12708, 2023. 1, 2

    2023 arXiv

  8. [16]

    Faster diffusion via temporal attention decomposition, 2024

    Haozhe Liu, Wentian Zhang, Jinheng Xie, Francesco Faccio, Mengmeng Xu, Tao Xiang, Mike Zheng Shou, Juan-Manuel Perez-Rua, and J ¨urgen Schmidhuber. Faster diffusion via temporal attention decomposition, 2024. 2

    2024

  9. [17]

    Flow straight and fast: Learning to generate and transfer data with rectified flow

    Xingchao Liu, Chengyue Gong, and qiang liu. Flow straight and fast: Learning to generate and transfer data with rectified flow. In The Eleventh International Conference on Learning Representations, 2023. 1, 2, 5

    2023

  10. [18]

    Dpm-solver: A fast ode solver for diffusion probabilistic model sampling in around 10 steps

    Cheng Lu, Yuhao Zhou, Fan Bao, Jianfei Chen, Chongxuan Li, and Jun Zhu. Dpm-solver: A fast ode solver for diffusion probabilistic model sampling in around 10 steps. Advances in Neural Information Processing Systems , 35:5775–5787, 2022

    2022

  11. [19]

    Dpm-solver++: Fast solver for guided sampling of diffusion probabilistic models

    Cheng Lu, Yuhao Zhou, Fan Bao, Jianfei Chen, Chongx- uan Li, and Jun Zhu. Dpm-solver++: Fast solver for guided sampling of diffusion probabilistic models. arXiv preprint arXiv:2211.01095, 2022. 1, 2

    2022 arXiv

  12. [20]

    Learning-to-cache: Accelerating diffusion trans- former via layer caching

    Xinyin Ma, Gongfan Fang, Michael Bi Mi, and Xinchao Wang. Learning-to-cache: Accelerating diffusion trans- former via layer caching. arXiv preprint arXiv:2406.01733,

    arXiv

  13. [21]

    Deepcache: Accelerating diffusion models for free

    Xinyin Ma, Gongfan Fang, and Xinchao Wang. Deepcache: Accelerating diffusion models for free. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 15762–15772, 2024. 2

    2024

  14. [22]

    Latte: Latent diffusion transformer for video generation

    Xin Ma, Yaohui Wang, Gengyun Jia, Xinyuan Chen, Zi- wei Liu, Yuan-Fang Li, Cunjian Chen, and Yu Qiao. Latte: Latent diffusion transformer for video generation. arXiv preprint arXiv:2401.03048, 2024. 1, 2

    2024 arXiv

  15. [23]

    The song describer dataset: a corpus of audio captions for music-and-language evaluation

    Ilaria Manco, Benno Weck, Seungheon Doh, Minz Won, Yixiao Zhang, Dmitry Bogdanov, Yusong Wu, Ke Chen, Philip Tovstogan, Emmanouil Benetos, Elio Quinton, Gy¨orgy Fazekas, and Juhan Nam. The song describer dataset: a corpus of audio captions for music-and-language evaluation. In...

    2023

  16. [24]

    Dit-3d: Exploring plain diffusion transformers for 3d shape generation

    Shentong Mo, Enze Xie, Ruihang Chu, Lanqing Hong, Matthias Niessner, and Zhenguo Li. Dit-3d: Exploring plain diffusion transformers for 3d shape generation. Advances in neural information processing systems , 36:67960–67971,

  17. [25]

    Null-text inversion for editing real im- ages using guided diffusion models

    Ron Mokady, Amir Hertz, Kfir Aberman, Yael Pritch, and Daniel Cohen-Or. Null-text inversion for editing real im- ages using guided diffusion models. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 6038–6047, 2023. 2

    2023

  18. [26]

    Scalable diffusion models with transformers

    William Peebles and Saining Xie. Scalable diffusion models with transformers. In Proceedings of the IEEE/CVF Inter- national Conference on Computer Vision, pages 4195–4205,

  19. [27]

    High-resolution image synthesis with latent diffusion models

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Bj ¨orn Ommer. High-resolution image synthesis with latent diffusion models. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 10684–10695, 2022. 2

    2022

  20. [28]

    Photorealistic text-to-image diffusion models with deep language understanding

    Chitwan Saharia, William Chan, Saurabh Saxena, Lala Li, Jay Whang, Emily L Denton, Kamyar Ghasemipour, Raphael Gontijo Lopes, Burcu Karagol Ayan, Tim Salimans, et al. Photorealistic text-to-image diffusion models with deep language understanding. Advances in neural information...

    2022

  21. [29]

    Progressive distillation for fast sampling of diffusion models

    Tim Salimans and Jonathan Ho. Progressive distillation for fast sampling of diffusion models. arXiv preprint arXiv:2202.00512, 2022. 1, 2

    2022 arXiv

  22. [30]

    Fora: Fast-forward caching in diffusion transformer acceleration

    Pratheba Selvaraju, Tianyu Ding, Tianyi Chen, Ilya Zharkov, and Luming Liang. Fora: Fast-forward caching in diffusion transformer acceleration. arXiv preprint arXiv:2407.01425 ,

    arXiv

  23. [31]

    Post-training quantization on diffusion models

    Yuzhang Shang, Zhihang Yuan, Bin Xie, Bingzhe Wu, and Yan Yan. Post-training quantization on diffusion models. In Proceedings of the IEEE/CVF conference on computer vi- sion and pattern recognition, pages 1972–1981, 2023. 1, 2

    1972

  24. [32]

    Deep unsupervised learning using nonequilibrium thermodynamics

    Jascha Sohl-Dickstein, Eric Weiss, Niru Maheswaranathan, and Surya Ganguli. Deep unsupervised learning using nonequilibrium thermodynamics. In International confer- ence on machine learning , pages 2256–2265. PMLR, 2015. 1, 2

    2015

  25. [33]

    Denois- ing diffusion implicit models

    Jiaming Song, Chenlin Meng, and Stefano Ermon. Denois- ing diffusion implicit models. In International Conference on Learning Representations, 2021. 2

    2021

  26. [34]

    Attention is all you need

    A Vaswani. Attention is all you need. Advances in Neural Information Processing Systems, 2017. 1, 2

    2017

  27. [35]

    Taming rectified flow for inversion and editing, 2024

    Jiangshan Wang, Junfu Pu, Zhongang Qi, Jiayi Guo, Yue Ma, Nisha Huang, Yuxin Chen, Xiu Li, and Ying Shan. Taming rectified flow for inversion and editing, 2024. 2

    2024

  28. [36]

    Vidprom: A million-scale real prompt-gallery dataset for text-to-video diffusion models

    Wenhao Wang and Yi Yang. Vidprom: A million-scale real prompt-gallery dataset for text-to-video diffusion models. In Thirty-eighth Conference on Neural Information Processing Systems, 2024. 6

    2024

  29. [37]

    Zhou Wang and A.C. Bovik. A universal image quality in- dex. IEEE Signal Processing Letters, 9(3):81–84, 2002. 5

    2002

  30. [38]

    Cache me if you can: Accel- erating diffusion models through block caching

    Felix Wimbauer, Bichen Wu, Edgar Schoenfeld, Xiaoliang Dai, Ji Hou, Zijian He, Artsiom Sanakoyeu, Peizhao Zhang, Sam Tsai, Jonas Kohler, et al. Cache me if you can: Accel- erating diffusion models through block caching. In Proceed- ings of the IEEE/CVF Conference on Computer V...

    2024

  31. [39]

    The unreasonable effectiveness of deep features as a perceptual metric

    Richard Zhang, Phillip Isola, Alexei A Efros, Eli Shecht- man, and Oliver Wang. The unreasonable effectiveness of deep features as a perceptual metric. In Proceedings of the IEEE conference on computer vision and pattern recogni- tion, pages 586–595, 2018. 5

    2018

  32. [40]

    Real-time video generation with pyramid attention broad- cast

    Xuanlei Zhao, Xiaolong Jin, Kai Wang, and Yang You. Real-time video generation with pyramid attention broad- cast. arXiv preprint arXiv:2408.12588, 2024. 2

    2024 arXiv

  33. [41]

    Open-sora: Democratizing efficient video production for all, 2024

    Zangwei Zheng, Xiangyu Peng, Tianji Yang, Chenhui Shen, Shenggui Li, Hongxin Liu, Yukun Zhou, Tianyi Li, and Yang You. Open-sora: Democratizing efficient video production for all, 2024. 1, 2 SmoothCache: A Universal Inference Acceleration Technique for Diffusion Transformers S...

    2024

  34. [42]

    3.3, the calibration sample size cho- sen does not dramatically change the generated schedule

    Ablations Addendum: Number of Calibra- tion Samples As mentioned in Sec. 3.3, the calibration sample size cho- sen does not dramatically change the generated schedule. Increasing the number of samples only affects the range of the confidence interval for the error curves, but ...

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.