Pith. sign in

REVIEW 4 major objections 6 minor 1 cited by

On Designing Diffusion Autoencoders for Efficient Generation and Representation Learning

T0 review · 4 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read The paper proposes DMZ, a diffusion autoencoder whose small binary latent is sampled from a fixed Bernoulli prior at inference, yielding efficient few-step generation and useful representations without auxiliary samplers or additional…

desk verdict DMZ is a solid empirical recipe for a diffusion autoencoder with binary latents and cross-attention, but the headline claim—no auxiliary sampler—only holds for tiny latent sizes, and the paper's own ablations show it. read the letter →

arxiv 2506.00136 v1 pith:6EQ727HJ submitted 2025-05-30 cs.LG

Magdalena Proszewska , Nikolay Malkin , N. Siddharth This is my paper
classification cs.LG
keywords diffusionautoencodersdiscretelatentvariablesrepresentationlearningefficientgenerationfew-stepsamplingBernoullipriorcross-attentionconditioningimage-to-imagetranslation
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 proposes DMZ, a diffusion autoencoder that uses a small discrete (binary) latent variable to condition the denoising process. Its central claim is that with the right design choices—binary latents, cross-attention conditioning, and a small latent size—the latent can be sampled at inference time from a fixed independent Bernoulli(0.5) prior, with no auxiliary sampler, no extra loss terms, and no constraints on the latent. The authors argue that this gives the best of both worlds: generation quality comparable to standard diffusion models at fewer denoising steps, and learned representations that support downstream classification, interpolation, and attribute editing. If correct, DMZ shows that efficient generation and representation learning can come from a single objective and a simple prior.

What carries the argument

The load-bearing mechanism is the small binary latent z, produced by a convolutional encoder and fed into the denoising UNet through cross-attention blocks, where z supplies the keys and values and the denoiser's own features supply the queries. Because |z| is small (16 for CIFAR-10, 64 for CelebA-64), the distribution of encoded codes averaged over the data stays close to an independent Bernoulli(0.5) product distribution, so the generative prior needs no training. The entire model is trained with the standard diffusion denoising loss, now conditioned on z, with no mutual-information penalty, no prior-fitting stage, and no auxiliary sampler at inference.

What would settle it

Train DMZ on CIFAR-10 with a 128-bit latent and compute the KL divergence between the data-averaged encoder distribution and the Bernoulli(0.5) product prior. If that divergence is large and FID at T=100 with Bernoulli-sampled z is substantially worse than the unconditional DDPM baseline, then the 'simple prior suffices' claim fails as |z| grows. A complementary test on more diverse data, such as ImageNet-64, would show whether the small-latent regime transfers beyond the three datasets used here.

Watch

Extended reading notes

Core claim

The paper's central discovery is that a diffusion autoencoder can function as both an efficient generator and a representation learner when its latent is binary, low-dimensional, and injected through cross-attention. The authors draw an analogy between conditioning a denoiser on a learned latent z and diffusion models with a learnable forward process: both carry input-dependent side information into the denoising trajectory, but the DA formulation avoids the terminal constraints that learned-noising models must satisfy. With |z| kept small, the data-averaged encoder distribution qφ(z|x0) is close enough to a product Bernoulli(0.5) prior that sampling z directly from that prior yields competitive FID scores, even at T=10 or T=20 steps. The same z serves as a representation for linear classification, bit-flip interpolation, and classifier-guided attribute edits, and two independently trained DMZ modules can be chained through a learned latent-space map for sketch-to-photo translation.

Load-bearing premise

The entire no-auxiliary-sampler claim rests on the assumption that the distribution of latent codes produced by the encoder, averaged over the data, is close enough to independent coin flips (Bernoulli(0.5)) to serve as a generative prior; the paper only keeps this true by choosing a very small latent size.

Editorial extensions

If this is right

  • DMZ achieves competitive FID scores with as few as 10-20 denoising steps on CIFAR-10 and CelebA-64, outperforming a same-architecture DDPM at those step counts.
  • Because z is sampled from a fixed Bernoulli prior, DMZ removes the auxiliary DDIM, GMM, or autoregressive samplers that previous diffusion autoencoders require for generation.
  • The latent representations enable downstream linear classification and attribute prediction at levels comparable to or better than InfoDiffusion, without any explicit regularization of z.
  • The framework extends to image-to-image translation: two independently trained DMZ models can be coupled by a learned MLP mapping between their latent spaces, as demonstrated on sketch-to-photo translation.
  • A pretrained unconditional DDPM can be finetuned into DMZ, offering faster training at some cost in few-step generation quality.

Reading between the lines

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

  • The DA-as-learned-noising analogy suggests a concrete diagnostic: fixing z and tracing the effective noise schedule of the denoiser should reveal the input-dependent 'shift' that z encodes; measuring this shift across t could make the connection quantitative.
  • The method's simplicity rests on small |z|, so a natural stress test is scaling to more diverse data: if ImageNet-scale data requires a larger latent to capture class structure, the Bernoulli prior may stop being adequate and the no-auxiliary-sampler property would break.
  • Since the encoder is trained purely by the denoising loss, DMZ provides a clean setting to ask which aspects of an image the diffusion loss forces into a discrete bottleneck; measuring the accuracy-vs-|z| trade-off could inform how much information such bottlenecks can carry.
  • The success of cross-attention over concatenation hints that how z touches the UNet matters more than the loss; this inductive-bias result could transfer to other conditioning-injection problems, such as class- or text-conditioning.
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

4 major / 6 minor

Summary. The paper proposes DMZ, a diffusion autoencoder variant that uses a small binary latent variable z, conditions the denoising U-Net on z through cross-attention, and is trained with only the standard diffusion denoising loss. The authors argue that this design connects diffusion autoencoders to diffusion models with learned forward processes, allowing efficient generation with few denoising steps while simultaneously learning representations useful for downstream classification, editing, interpolation, and image-to-image translation. Experiments are reported on CIFAR-10, CelebA-64, Edges2Handbags, and CelebA-HQ finetuning, with ablations over latent type, conditioning method, latent size, and sampling strategy. The main claims are that DMZ outperforms standard DDPM baselines in few-step generation, matches or exceeds prior diffusion autoencoders in representation quality without auxiliary losses or samplers, and extends to multimodal translation.

Significance. If the claims hold, DMZ is a simple and useful architecture: it unifies generation and representation learning in one diffusion model, requires no auxiliary prior model in its advertised small-latent regime, and the paper includes several strengths: extensive ablations, qualitative demonstrations of interpolation and editing, finetuning experiments from pretrained DDPMs, and public code. The central novelty is the combination of binary latents, cross-attention conditioning, and explicit reliance on a fixed Bernoulli prior, which is an elegant way to avoid auxiliary samplers. However, the significance is moderated by the fact that the no-auxiliary-sampler property is established only for hand-picked small latent sizes, exactly the regime where representation quality is weakest, and by the lack of any direct measurement of the match between the aggregate posterior qφ(z|x0) and the Bernoulli prior. The multimodal extension explicitly falls back to PixelSNAIL, which is in tension with the paper's central claim. These issues are fixable with scoping and additional analysis, so the paper has clear potential, but the current form overstates the generality of its main result.

major comments (4)
  1. [§4.5 (Table 6) and Appendix A.3] The central contribution 'no auxiliary samplers' (Section 5) is not supported outside the small-|z| regime. Table 6 shows that DMZ-16 on CIFAR-10 with |z|=128, Bernoulli sampling gives FID 17.23 at T=100, versus 4.46 when z is drawn from the data posterior; an autoregressive PixelSNAIL prior recovers only 9.54. Appendix A.3 states that in the multimodal setting with |z|=512, 'we use PixelSNAIL for unconditional generation, as the latent size |z|=512 was chosen to optimise reconstruction loss rather than efficient sampling of z.' Since Table 2 shows representation quality improves with |z| (39.5 to 45.6 accuracy on CIFAR-10 and 79.4 to 81.0 AUROC on CelebA-64), the regime of strongest representations is exactly the regime where the Bernoulli prior is inadequate. The paper should either measure the divergence between qφ(z|x0) and the Bernoulli prior (e.g., marginal statistics or FID with a learned prior) and report such a check for the CelebA-64 configuration used in Table 1, or explicitly scope the no-auxiliary-sampler claim to the small-|z| unimodal setting.
  2. [§4.2, Table 1] The FID comparisons in Table 1 mix evaluation sizes: the table reports 10K-sample FID for DMZ and DDPM, while models marked * use 50K samples. The text further states that DMZ achieves FID 2.83 on CIFAR-10 at T=100 with 50K samples, a number not in the table. Because FID is sensitive to the number of samples, direct comparisons across columns with different evaluation sizes are not apples-to-apples. Please report all numbers in a given comparison at the same evaluation size, or split into separate tables.
  3. [§4, Limitations] All headline numbers, including Tables 1, 2, 3, 6, and 7, come from single training runs. The Limitations paragraph acknowledges this ('repeated runs would improve statistical confidence'), but the abstract and Section 5 make unqualified comparative claims such as 'outperforming existing diffusion autoencoders.' Given the small differences between DMZ and baselines in several rows of Table 1, single-run results do not establish these claims. Please provide multiple seeds for the key generation and representation comparisons, or at minimum report bootstrap/error estimates and qualify the comparative wording.
  4. [§3 and Algorithm 1] The training procedure is underspecified: Algorithm 1 says 'Extract relaxed code z from x0 via encoder parametrized by φ,' but the paper never defines the encoder output distribution, how a binary sample is obtained from a relaxed code, or how gradients flow through the sampling step. This matters because the generative claim rests on sampling z from a Bernoulli(0.5) prior at inference, while the training objective Eq. (5) uses z from qφ(z|x0). If training uses a continuous relaxation and inference binarizes, the mismatch should be described and analyzed; if the encoder outputs Bernoulli probabilities, the text should say so. Please add the precise parameterization of qφ and the straight-through/relaxation scheme.
minor comments (6)
  1. [§4.2] Thess scores should be These scores.
  2. [§4.3] dataseset should be dataset.
  3. [§4.4, Table 3] DMZ has MSE 0.209, far higher than DDBM's 0.013-0.040; the claim of competitive performance should be qualified with this large gap.
  4. [Table 4] The column arrangement is confusing: the header 'prior qφ(z|x0)' appears to apply only to the Bernoulli row.
  5. [§3, Choice of latent z] The choice of p=0.5 for the Bernoulli prior is not ablated or justified; a small experiment varying p would strengthen the claim.
  6. [§5] The phrase 'additional, input-dependent priors' in the conclusion is vague; the paper does not formally introduce a prior family beyond the fixed Bernoulli distribution.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: DMZ's claims are empirical and explicitly conditional on small latent size; no derivation reduces to its inputs.

full rationale

The paper's central claims are empirical demonstrations rather than derivations. DMZ is trained with the standard diffusion loss (Eq. 5) conditioned on a binary latent z, and the paper evaluates FID, NLL, accuracy, and AUROC against external baselines (DDPM, DiffAE, InfoDiffusion, etc.). No equation defines a predicted quantity in terms of the claim itself, and no parameter is fitted to a subset of data and then reported as a prediction. The closest edge case is the 'no auxiliary sampler' claim in Section 5 and Algorithm 2: it holds only because the latent dimensionality is deliberately kept small, as the paper openly states ('The dimensionality of z was deliberately kept small...'; 'we adopt low-dimensional z, optimising for direct sampling'). Table 6 confirms that Bernoulli sampling degrades at larger |z| (FID 4.79 to 17.23 for |z|=16 to 128 at T=100), and Appendix A.3 admits the use of PixelSNAIL for |z|=512. This is a real scope limitation and a verification gap—the aggregate posterior qφ(z|x0) is never directly compared to Bernoulli(0.5)—but it is not circularity: the design choice is disclosed, the claim is explicitly conditional ('In lower-dimensional latent spaces, the model better leverages the prior'), and the paper does not present the no-auxiliary-sampler property as a forced consequence of a definition or a fitted parameter. The cited prior works are external (DiffAE, InfoDiffusion, VDM, etc.), and there is no load-bearing self-citation chain. The mathematical identities in Eqs. 1-5 are standard diffusion-model equations with z as an additional conditioning input; none of them reduce to the paper's conclusions. Under the strict definition of circularity, the paper is self-contained, so the honest finding is no significant circularity.

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

This is an empirical paper, so the ledger captures the design hyperparameters and unstated modeling premises that the conclusions depend on, rather than fitted constants in a derivation.

free parameters (3)
  • Latent dimensionality |z| = 16/32/64 (CIFAR-10), 64/128/256 (CelebA-64), 512 (Edges2Handbags)
    Chosen per dataset 'guided by the number of available labels' (§4) and by reconstruction MSE for multimodal (§A.3). The claim that direct Bernoulli sampling works depends critically on keeping |z| small; Table 6 shows FID degrades sharply for |z|≥128.
  • Bernoulli prior mean p = 0.5
    Uniform Bernoulli assumed as the latent prior (Algorithm 2). Not fitted, but its adequacy is a load-bearing assumption; the paper does not check aggregate posterior statistics.
  • MLP mapping depth L for γ = 6
    Selected by MSE among L∈{1,2,4,6,8} (§A.3). This hyperparameter is fitted to the same validation metric used in the main comparison.
assumptions (4)
  • domain assumption The denoising objective (Eq. 5) with a fixed Gaussian noise schedule defines a valid generative model when z is sampled from a prior.
    Assumed throughout; standard for DAs, but the paper relies on it to claim that Bernoulli(0.5) sampling yields a proper generative model.
  • domain assumption The learned encoder qφ(z|x0) and denoiser ϵθ(xt,t,z) trained jointly via Eq. (5) do not collapse z to an uninformative constant.
    Section 3 'Learning with latent z' argues redundancy, but no proof. The MI measurement in §4.1 attempts to show informativeness, but it is a heuristic.
  • ad hoc to paper Cross-attention with z as keys/values is a strictly better conditioning mechanism than concatenating z with the time embedding.
    Justified by a single ablation (Table 5) on CIFAR-10; no theoretical or multi-dataset support.
  • domain assumption FID and downstream linear classification are valid proxies for generation quality and representation quality.
    Used in all comparisons; standard in the field, but the paper's conclusions rest on these metrics.

how reviews work

0 comments
Cite this review

Pith. "Pith review of On Designing Diffusion Autoencoders for Efficient Generation and Representation Learning." pith.science (2026). https://pith.science/paper/6EQ727HJ

@misc{pith2026250600136,
  author       = {Pith},
  title        = {Pith review of: On Designing Diffusion Autoencoders for Efficient Generation and Representation Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6EQ727HJ}},
  note         = {Machine review of arXiv:2506.00136}
}
read the original abstract

Diffusion autoencoders (DAs) are variants of diffusion generative models that use an input-dependent latent variable to capture representations alongside the diffusion process. These representations, to varying extents, can be used for tasks such as downstream classification, controllable generation, and interpolation. However, the generative performance of DAs relies heavily on how well the latent variables can be modelled and subsequently sampled from. Better generative modelling is also the primary goal of another class of diffusion models -- those that learn their forward (noising) process. While effective at adjusting the noise process in an input-dependent manner, they must satisfy additional constraints derived from the terminal conditions of the diffusion process. Here, we draw a connection between these two classes of models and show that certain design decisions (latent variable choice, conditioning method, etc.) in the DA framework -- leading to a model we term DMZ -- allow us to obtain the best of both worlds: effective representations as evaluated on downstream tasks, including domain transfer, as well as more efficient modelling and generation with fewer denoising steps compared to standard DMs.

Figures

Figures reproduced from arXiv: 2506.00136 by the authors.

Figure 1
Figure 1. Top: Basic diffusion reverses a Markovian noising process from xT (possibly via predicted x0 at each step). Middle: A generalisation where gener￾ation reverses a non-Markovian learned noising process, marginalising out un￾known x0. Bottom: DMZ, where gen￾eration conditions on latent z. Solid and dashed arrows denote noising and gen￾eration respectively. Red arrows denote learned parametric models. Blue objects denot… view at source ↗
Figure 2
Figure 2. Top: Denoiser block with time con￾ditioning and optional attention. Mid, bottom: Two conditioning strategies with z. Conditioning on latent z: Given a latent z, another key design decision for effective modelling is what form conditioning the denoiser takes. In the standard case, the denoiser simply takes the noisy observation xt along with indication of time step t to predict the denoised data xt−1, possibly via pr… view at source ↗
Figure 3
Figure 3. Comparison of training curves for DMZ and the baseline DDPM. Dashed lines correspond [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (12 more)
Figure 4
Figure 4. Figure 4: Mutual information between the representations z learned by DMZ￾16 on CIFAR-10 and xt from the nois￾ing process (blue) or from the de￾noising process (orange; starting from {xT , z}, xT ∼ N (0, I)). On the denoising process: Next, we examine the role of z during differ…
Figure 5
Figure 5. Figure 5: Examplar generations from z of a single image by DMZ, with rows corresponding to |z| = 64, 128, 256. Next, we illustrate the properties of the learned representa￾tions z through interpolation examples, where transitions between latent vectors lead to gradual changes in…
Figure 6
Figure 6. Figure 6: Qualitative evaluation on CelebA-64. images from the dataset and extract their corresponding encodings, zsource and ztarget. We then perform discrete interpolations by sequentially flipping bits in zsource to match those in ztarget. Fig. 6a presents examples for DMZ-25…
Figure 7
Figure 7. Figure 7: Examples of sketch-to-photo translations [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]
Figure 10
Figure 10. Figure 10: The reconstruction error of DMZ models trained on Handbags-64 measured on 10K images, an upperbound for Edges2Handbags task. DMZ modules We use two instances of DMZ-512: one trained on Edges-64 and the other on Handbags-64. In the sketch-to-photo task, only the model …
Figure 8
Figure 8. Figure 8: Examples of discrete interpolations between codes [PITH_FULL_IMAGE:figures/full_fig_p016_8.png]
Figure 9
Figure 9. Figure 9: Examples of classifier-based edits for T = 100 on CelebA-64 using DMZ-256. For the first image x0, we change following CelebA attributes: glasses, male, hat; for the second: gray hair, bald, smile; the third: bangs, blond hair, hat; the fourth: bangs, male, earrings; t…
Figure 11
Figure 11. Figure 11: Qualitative results showing additional capabilities of the DMZ image-to-image framework. [PITH_FULL_IMAGE:figures/full_fig_p017_11.png]
Figure 12
Figure 12. Figure 12: Evolution of representation quality throughout training. [PITH_FULL_IMAGE:figures/full_fig_p018_12.png]
Figure 13
Figure 13. Figure 13: Comparison of representations learned by DMZ on CIFAR-10 and CelebA-64 for varying [PITH_FULL_IMAGE:figures/full_fig_p019_13.png]
Figure 14
Figure 14. Figure 14: Images generated with varying numbers of denoising steps [PITH_FULL_IMAGE:figures/full_fig_p020_14.png]
Figure 15
Figure 15. Figure 15: Generated sample images produced using T = 100 diffusion steps. 21 [PITH_FULL_IMAGE:figures/full_fig_p021_15.png]

Discussion (0). Sign in 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. When Geometry Aligns: Dihedral Hidden-State Transformations in UNet, ViT, and DiT Architectures

    cs.LG 2026-07 conditional novelty 6.0 of 10

    Geometrically consistent dihedral flips of hidden states stabilize U-Net, ViT, and DiT computation; inconsistent flips produce architecture-specific mismatch and drift.

Reference graph

Works this paper leans on

50 extracted references · 32 canonical work pages · cited by 1 Pith paper

  1. [1]

    Vetrov, and Christian Andersson Naesseth

    Grigory Bartosh, Dmitry P. Vetrov, and Christian Andersson Naesseth. Neural flow diffusion models: Learnable forward process for improved diffusion modelling. In Amir Globersons, Lester Mackey, Danielle Belgrave, Angela Fan, Ulrich Paquet, Jakub M. Tomczak, and Cheng Zhang, editors, Advances in Neural Information Processing Systems 38: Annual Conference o...

    work page 2024

  2. [2]

    Mine: Mutual information neural estimation, 2018

    Mohamed Ishmael Belghazi, Aristide Baratin, Sai Rajeswar, Sherjil Ozair, Yoshua Bengio, Aaron Courville, and R Devon Hjelm. Mine: Mutual information neural estimation, 2018. URL https://arxiv.org/abs/1801.04062

    arXiv 2018

  3. [3]

    Representation learning: A review and new perspectives, 2014

    Yoshua Bengio, Aaron Courville, and Pascal Vincent. Representation learning: A review and new perspectives, 2014. URL https://arxiv.org/abs/1206.5538

    arXiv 2014

  4. [4]

    Pixelsnail: An improved autoregressive generative model

    Xi Chen, Nikhil Mishra, Mostafa Rohaninejad, and Pieter Abbeel. Pixelsnail: An improved autoregressive generative model. In Jennifer G. Dy and Andreas Krause, editors, Proceedings of the 35th International Conference on Machine Learning, ICML 2018, Stockholmsmäs- san, Stockholm, Sweden, July 10-15, 2018 , volume 80 of Proceedings of Machine Learn- ing Res...

    work page 2018

  5. [5]

    Diffusion models beat gans on image syn- thesis

    Prafulla Dhariwal and Alexander Quinn Nichol. Diffusion models beat gans on image syn- thesis. In Marc’Aurelio Ranzato, Alina Beygelzimer, Yann N. Dauphin, Percy Liang, and Jennifer Wortman Vaughan, editors,Advances in Neural Information Processing Systems 34: Annual Conference on Neural Information Processing Systems 2021, NeurIPS 2021, December 6-14, 20...

    work page 2021

  6. [6]

    Denoising diffusion probabilistic models

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising diffusion probabilistic models. In Hugo Larochelle, Marc’Aurelio Ranzato, Raia Hadsell, Maria-Florina Balcan, and Hsuan- Tien Lin, editors, Advances in Neural Information Processing Systems 33: Annual Con- ference on Neural Information Processing Systems 2020, NeurIPS 2020, December 6- 12, 2020, virtual...

    work page 2020

  7. [7]

    Hudson, Daniel Zoran, Mateusz Malinowski, Andrew K

    Drew A. Hudson, Daniel Zoran, Mateusz Malinowski, Andrew K. Lampinen, Andrew Jaegle, James L. McClelland, Loic Matthey, Felix Hill, and Alexander Lerchner. SODA: bottleneck diffusion models for representation learning. In IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2024, Seattle, WA, USA, June 16-22, 2024, pages 23115–23127. IEEE,...

    arXiv 2024

  8. [8]

    Phillip Isola, Jun-Yan Zhu, Tinghui Zhou, and Alexei A. Efros. Image-to-image translation with conditional adversarial networks. In 2017 IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2017, Honolulu, HI, USA, July 21-26, 2017 , pages 5967–5976. IEEE Computer Society, 2017. doi: 10.1109/CVPR.2017.632. URL https://doi.org/10. 1109/CVPR.2017.632

    doi:10.1109/cvpr.2017.632 2017

Show all 50 references
  1. [9]

    Kingma and Max Welling

    Diederik P. Kingma and Max Welling. Auto-encoding variational bayes. In Yoshua Bengio and Yann LeCun, editors, 2nd International Conference on Learning Representations, ICLR 2014, Banff, AB, Canada, April 14-16, 2014, Conference Track Proceedings , 2014. URL http://arxiv.org/a...

    2014 arXiv

  2. [10]

    Kingma, Tim Salimans, Ben Poole, and Jonathan Ho

    Diederik P. Kingma, Tim Salimans, Ben Poole, and Jonathan Ho. Variational diffusion models,

  3. [11]

    Learning multiple layers of features from tiny images

    Alex Krizhevsky. Learning multiple layers of features from tiny images. Technical report, University of Toronto, 2009. 10

    2009

  4. [12]

    Anna Kuzina and Jakub M. Tomczak. Hierarchical vae with a diffusion-based vampprior, 2024. URL https://arxiv.org/abs/2412.01373

    2024 arXiv

  5. [13]

    Priorgrad: Improving conditional denoising diffusion models with data-dependent adaptive prior

    Sang-gil Lee, Heeseung Kim, Chaehun Shin, Xu Tan, Chang Liu, Qi Meng, Tao Qin, Wei Chen, Sungroh Yoon, and Tie-Yan Liu. Priorgrad: Improving conditional denoising diffusion models with data-dependent adaptive prior. In The Tenth International Conference on Learning Representat...

    2022

  6. [14]

    Theodorou, Weili Nie, and Anima Anandkumar

    Guan-Horng Liu, Arash Vahdat, De-An Huang, Evangelos A. Theodorou, Weili Nie, and Anima Anandkumar. I 2sb: Image-to-image schrödinger bridge, 2023. URL https://arxiv.org/ abs/2302.05872

    2023 arXiv

  7. [15]

    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, ICLR 2023, Kigali, Rwanda, May 1-5, 2023. OpenReview.net, 2023. URL https://ope...

    2023

  8. [16]

    Deep learning face attributes in the wild

    Ziwei Liu, Ping Luo, Xiaogang Wang, and Xiaoou Tang. Deep learning face attributes in the wild. In 2015 IEEE International Conference on Computer Vision, ICCV 2015, Santiago, Chile, December 7-13, 2015, pages 3730–3738. IEEE Computer Society, 2015. doi: 10.1109/ICCV . 2015.425...

    2015 doi

  9. [17]

    Decoupled weight decay regularization

    Ilya Loshchilov and Frank Hutter. Decoupled weight decay regularization. In 7th International Conference on Learning Representations, ICLR 2019, New Orleans, LA, USA, May 6-9, 2019. OpenReview.net, 2019. URL https://openreview.net/forum?id=Bkg6RiCqY7

    2019

  10. [18]

    Sdedit: Guided image synthesis and editing with stochastic differential equations

    Chenlin Meng, Yutong He, Yang Song, Jiaming Song, Jiajun Wu, Jun-Yan Zhu, and Stefano Ermon. Sdedit: Guided image synthesis and editing with stochastic differential equations. In The Tenth International Conference on Learning Representations, ICLR 2022, Virtual Event, April 25...

    2022

  11. [19]

    Discrete sequential prediction of continuous actions for deep rl, 2017

    Luke Metz, Julian Ibarz, Navdeep Jaitly, and James Davidson. Discrete sequential prediction of continuous actions for deep rl, 2017. URL https://arxiv.org/abs/1705.05035

    2017 arXiv

  12. [20]

    Edan Jacob Meyer, Adam White, and Marlos C. Machado. Harnessing discrete representations for continual reinforcement learning. Reinforcement Learning Journal, 2:606–628, 2024

    2024

  13. [21]

    Diamos, Erich Elsen, David García, Boris Ginsburg, Michael Houston, Oleksii Kuchaiev, Ganesh Venkatesh, and Hao Wu

    Paulius Micikevicius, Sharan Narang, Jonah Alben, Gregory F. Diamos, Erich Elsen, David García, Boris Ginsburg, Michael Houston, Oleksii Kuchaiev, Ganesh Venkatesh, and Hao Wu. Mixed precision training. In 6th International Conference on Learning Representations, ICLR 2018, Va...

    2018

  14. [22]

    Improved denoising diffusion probabilistic models

    Alexander Quinn Nichol and Prafulla Dhariwal. Improved denoising diffusion probabilistic models. In Marina Meila and Tong Zhang, editors, Proceedings of the 38th International Conference on Machine Learning, ICML 2021, 18-24 July 2021, Virtual Event, volume 139 of Proceedings ...

    2021

  15. [23]

    Diffenc: Variational diffusion with a learned encoder

    Beatrix Miranda Ginn Nielsen, Anders Christensen, Andrea Dittadi, and Ole Winther. Diffenc: Variational diffusion with a learned encoder. In The Twelfth International Conference on Learning Representations, ICLR 2024, Vienna, Austria, May 7-11, 2024. OpenReview.net, 2024. URL ...

    2024

  16. [24]

    Input perturbation reduces exposure bias in diffusion models

    Mang Ning, Enver Sangineto, Angelo Porrello, Simone Calderara, and Rita Cucchiara. Input perturbation reduces exposure bias in diffusion models. In Andreas Krause, Emma Brunskill, Kyunghyun Cho, Barbara Engelhardt, Sivan Sabato, and Jonathan Scarlett, editors,International Con...

    2023

  17. [25]

    Andrey Okhotin, Dmitry Molchanov, Vladimir Arkhipkin, Grigory Bartosh, Viktor Ohanesian, Aibek Alanov, and Dmitry P. Vetrov. Star-shaped denoising diffusion probabilistic models. In Alice Oh, Tristan Naumann, Amir Globerson, Kate Saenko, Moritz Hardt, and Sergey Levine, editor...

    2023

  18. [26]

    Diffusevae: Efficient, controllable and high-fidelity generation from low-dimensional latents, 2022

    Kushagra Pandey, Avideep Mukherjee, Piyush Rai, and Abhishek Kumar. Diffusevae: Efficient, controllable and high-fidelity generation from low-dimensional latents, 2022. URL https: //arxiv.org/abs/2201.00308

    2022 arXiv

  19. [27]

    Pytorch: An imperative style, high-performance deep learning library

    Adam Paszke, Sam Gross, Francisco Massa, Adam Lerer, James Bradbury, Gregory Chanan, Trevor Killeen, Zeming Lin, Natalia Gimelshein, Luca Antiga, Alban Desmaison, Andreas Köpf, Edward Yang, Zachary DeVito, Martin Raison, Alykhan Tejani, Sasank Chilamkurthy, Benoit Steiner, Lu ...

    2019

  20. [28]

    Diffusion autoencoders: Toward a meaningful and decodable representation

    Konpat Preechakul, Nattanat Chatthee, Suttisak Wizadwongsa, and Supasorn Suwajanakorn. Diffusion autoencoders: Toward a meaningful and decodable representation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pages 10619–10629, 2022

    2022

  21. [29]

    Generating diverse high-fidelity images with VQ-V AE-2

    Ali Razavi, Aäron van den Oord, and Oriol Vinyals. Generating diverse high-fidelity images with VQ-V AE-2. In Hanna M. Wallach, Hugo Larochelle, Alina Beygelzimer, Florence d’Alché-Buc, Emily B. Fox, and Roman Garnett, editors, Advances in Neu- ral Information Processing Syste...

    2019

  22. [30]

    Discrete variational autoencoders

    Jason Tyler Rolfe. Discrete variational autoencoders. In 5th International Conference on Learn- ing Representations, ICLR 2017, Toulon, France, April 24-26, 2017, Conference Track Proceed- ings. OpenReview.net, 2017. URL https://openreview.net/forum?id=ryMxXPFex

    2017

  23. [31]

    pytorch-fid: FID Score for PyTorch

    Maximilian Seitzer. pytorch-fid: FID Score for PyTorch. https://github.com/mseitzer/ pytorch-fid, 2020. Version 0.3.0

    2020

  24. [32]

    Denoising diffusion implicit models

    Jiaming Song, Chenlin Meng, and Stefano Ermon. Denoising diffusion implicit models. In 9th International Conference on Learning Representations, ICLR 2021, Virtual Event, Aus- tria, May 3-7, 2021. OpenReview.net, 2021. URL https://openreview.net/forum?id= St1giarCHLP

    2021

  25. [33]

    Kingma, Abhishek Kumar, Stefano Ermon, and Ben Poole

    Yang Song, Jascha Sohl-Dickstein, Diederik P. Kingma, Abhishek Kumar, Stefano Ermon, and Ben Poole. Score-based generative modeling through stochastic differential equations. In 9th In- ternational Conference on Learning Representations, ICLR 2021, Virtual Event, Austria, May ...

    2021

  26. [34]

    Dual diffusion implicit bridges for image-to-image translation

    Xuan Su, Jiaming Song, Chenlin Meng, and Stefano Ermon. Dual diffusion implicit bridges for image-to-image translation. In The Eleventh International Conference on Learning Rep- resentations, ICLR 2023, Kigali, Rwanda, May 1-5, 2023 . OpenReview.net, 2023. URL https://openrevi...

    2023

  27. [35]

    NV AE: A deep hierarchical variational autoencoder

    Arash Vahdat and Jan Kautz. NV AE: A deep hierarchical variational autoencoder. In Hugo Larochelle, Marc’Aurelio Ranzato, Raia Hadsell, Maria-Florina Balcan, and Hsuan- Tien Lin, editors, Advances in Neural Information Processing Systems 33: Annual Con- ference on Neural Infor...

    2020

  28. [36]

    Macready, Zhengbing Bian, Amir Khoshaman, and Evgeny Andriyash

    Arash Vahdat, William G. Macready, Zhengbing Bian, Amir Khoshaman, and Evgeny Andriyash. DV AE++: discrete variational autoencoders with overlapping transformations. In Jennifer G. Dy and Andreas Krause, editors, Proceedings of the 35th International Conference on Machine Lear...

    2018

  29. [37]

    Neural discrete representation learning

    Aäron van den Oord, Oriol Vinyals, and Koray Kavukcuoglu. Neural discrete representation learning. In Isabelle Guyon, Ulrike von Luxburg, Samy Bengio, Hanna M. Wallach, Rob Fergus, S. V . N. Vishwanathan, and Roman Garnett, editors,Advances in Neural Information Processing Sys...

    2017

  30. [38]

    Infodiffusion: Representation learning using information maximizing diffusion models

    Yingheng Wang, Yair Schiff, Aaron Gokaslan, Weishen Pan, Fei Wang, Christopher De Sa, and V olodymyr Kuleshov. Infodiffusion: Representation learning using information maximizing diffusion models. In Andreas Krause, Emma Brunskill, Kyunghyun Cho, Barbara Engelhardt, Sivan Saba...

    2023

  31. [39]

    Binary latent diffusion

    Ze Wang, Jiang Wang, Zicheng Liu, and Qiang Qiu. Binary latent diffusion. In IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2023, Vancouver, BC, Canada, June 17-24, 2023, pages 22576–22585. IEEE, 2023. doi: 10.1109/CVPR52729.2023.02162. URL https://doi.or...

    2023

  32. [41]

    Disdiff: Unsupervised disentanglement of diffusion probabilistic models

    Tao Yang, Yuwang Wang, Yan Lu, and Nanning Zheng. Disdiff: Unsupervised disentanglement of diffusion probabilistic models. In Alice Oh, Tristan Naumann, Amir Globerson, Kate Saenko, Moritz Hardt, and Sergey Levine, editors, Advances in Neural Information Processing Systems 36:...

    2023

  33. [42]

    Diffusion model as representation learner

    Xingyi Yang and Xinchao Wang. Diffusion model as representation learner. In IEEE/CVF International Conference on Computer Vision, ICCV 2023, Paris, France, October 1-6, 2023, pages 18892–18903. IEEE, 2023. doi: 10.1109/ICCV51070.2023.01736. URL https://doi. org/10.1109/ICCV510...

    2023

  34. [43]

    Exploring diffusion time-steps for unsupervised representation learning

    Zhongqi Yue, Jiankun Wang, Qianru Sun, Lei Ji, Eric I-Chao Chang, and Hanwang Zhang. Exploring diffusion time-steps for unsupervised representation learning. In The Twelfth Interna- tional Conference on Learning Representations, ICLR 2024, Vienna, Austria, May 7-11, 2024. Open...

    2024

  35. [44]

    Efros, Eli Shechtman, and Oliver Wang

    Richard Zhang, Phillip Isola, Alexei A. Efros, Eli Shechtman, and Oliver Wang. The un- reasonable effectiveness of deep features as a perceptual metric. In 2018 IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2018, Salt Lake City, UT, USA, June 18-22, 2018 , p...

    2018 doi

  36. [45]

    Unsupervised representation learning from pre-trained diffusion probabilistic models

    Zijian Zhang, Zhou Zhao, and Zhijie Lin. Unsupervised representation learning from pre-trained diffusion probabilistic models. In Sanmi Koyejo, S. Mohamed, A. Agar- wal, Danielle Belgrave, K. Cho, and A. Oh, editors, Advances in Neural Infor- mation Processing Systems 35: Annu...

    2022

  37. [46]

    Unsupervised discovery of interpretable directions in h-space of pre-trained diffusion models, 2023

    Zijian Zhang, Luping Liu, Zhijie Lin, Yichen Zhu, and Zhou Zhao. Unsupervised discovery of interpretable directions in h-space of pre-trained diffusion models, 2023. URL https: //arxiv.org/abs/2310.09912

    2023 arXiv

  38. [47]

    Shiftddpms: Exploring conditional diffusion models by shifting diffusion trajectories

    Zijian Zhang, Zhou Zhao, Jun Yu, and Qi Tian. Shiftddpms: Exploring conditional diffusion models by shifting diffusion trajectories. In Brian Williams, Yiling Chen, and Jennifer Neville, editors, Thirty-Seventh AAAI Conference on Artificial Intelligence, AAAI 2023, Thirty-Fift...

    2023

  39. [48]

    Infovae: Information maximizing variational autoencoders, 2017

    Shengjia Zhao, Jiaming Song, and Stefano Ermon. Infovae: Information maximizing variational autoencoders, 2017. URL https://arxiv.org/abs/1706.02262

    2017 arXiv

  40. [49]

    Denoising diffusion bridge models

    Linqi Zhou, Aaron Lou, Samar Khanna, and Stefano Ermon. Denoising diffusion bridge models. In The Twelfth International Conference on Learning Representations, ICLR 2024, Vienna, Austria, May 7-11, 2024. OpenReview.net, 2024. URL https://openreview.net/forum? id=FKksTayvGo

    2024

  41. [50]

    Discrete autoencoders for sequence models, 2018

    Łukasz Kaiser and Samy Bengio. Discrete autoencoders for sequence models, 2018. URL https://arxiv.org/abs/1801.09797. 14 A Technical Appendices and Supplementary Material A.1 DMZ in relation to prior work Table 8 shows how DMZ relates to prior work on diffusion autoencoders. A...

    2018 arXiv

  42. [2021]

    URL https://arxiv.org/abs/2107.00630

    arXiv

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.