pith. machine review for the scientific record. sign in

arxiv: 2511.08416 · v3 · submitted 2025-11-11 · 📡 eess.SP · cs.IT· cs.LG· cs.MM· math.IT

Generative AI Meets 6G and Beyond: Diffusion Models for Semantic Communications

Hai-Long Qin , Jincheng Dai , Guo Lu , Shuo Shao , Sixian Wang , Tongda Xu , Wenjun Zhang , Ping Zhang
show 1 more author
Khaled B. Letaief
This is my paper

Pith reviewed 2026-05-17 23:36 UTC · model grok-4.3

classification 📡 eess.SP cs.ITcs.LGcs.MMmath.IT
keywords diffusion modelssemantic communicationsgenerative AI6G networkswireless communicationsinverse problemsconditional generationposterior inference
0
0 comments X

The pith

Diffusion models can serve as foundational engines for semantic communications in 6G and beyond by turning minimal meaning cues into full reconstructions.

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

The paper argues that wireless systems are nearing capacity limits, so shifting from exact bit transmission to conveying meaning lets receivers generate content using learned priors instead of receiving every detail. Diffusion models stand out among generative AI tools because of their high-quality outputs, stable training, and solid theoretical base. The authors supply a tutorial that covers score-based foundations plus three pillars of conditional diffusion for control, efficient diffusion for speed, and generalized diffusion for adaptation across domains. They also recast semantic decoding as an inverse problem to connect the field with computational imaging techniques. This setup would matter if true because it supports extreme compression while holding semantic fidelity in human, machine, and agent scenarios.

Core claim

The paper claims that diffusion models, thanks to their generation quality and theoretical foundations, provide a systematic way to implement generative semantic communications through score-based methods, conditional controllable generation, accelerated inference techniques, cross-domain adaptations, and an inverse problem view that treats semantic decoding as posterior inference.

What carries the argument

Score-based diffusion models that reverse a gradual noising process to produce samples, applied here to reconstruct semantically faithful content from sparse cues sent over wireless channels.

If this is right

  • Semantic communications achieve much higher compression ratios while keeping meaning intact for human-centric uses such as media delivery.
  • Machine-centric tasks gain accurate reconstructions suited to specific objectives through conditioned diffusion.
  • Agent-centric coordination in networks benefits from robust generative priors that handle channel variations.
  • The inverse problem framing allows borrowing reconstruction methods from imaging to improve decoding under impairments.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • System designers could build standard protocols around semantic encoders paired with diffusion decoders rather than traditional bit pipelines.
  • Real deployments might require channel-aware fine-tuning or hybrid models to meet latency targets in live networks.
  • Similar generative techniques could extend to other resource-constrained settings like edge computing or sensor networks.

Load-bearing premise

Diffusion models trained on general data can be conditioned and adapted to preserve semantic fidelity across varied communication scenarios and channel conditions without major extra training or custom losses.

What would settle it

Measurements showing that diffusion-based reconstructions lose essential meaning or become unstable under realistic wireless noise, fading, or interference would indicate the approach does not reliably support semantic communications.

Figures

Figures reproduced from arXiv: 2511.08416 by Guo Lu, Hai-Long Qin, Jincheng Dai, Khaled B. Letaief, Ping Zhang, Shuo Shao, Sixian Wang, Tongda Xu, Wenjun Zhang.

Figure 1
Figure 1. Figure 1: Schematic diagram of Weaver’s three-level communication [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Statistical results manifesting the rapid development of [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Overview of the article organization. Each colored box represents a major section of the article. [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Comparison between discriminative and generative modeling [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Score-based modeling pipeline for diffusion models. (a) Score [PITH_FULL_IMAGE:figures/full_fig_p006_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Forward-reverse SDE pipeline for score-based diffusion models. The forward SDE progressively corrupts the input [PITH_FULL_IMAGE:figures/full_fig_p007_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Solving probability flow ODEs with the Predictor-Corrector [PITH_FULL_IMAGE:figures/full_fig_p008_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Underlying mechanism of flow-matching process. [PITH_FULL_IMAGE:figures/full_fig_p013_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Generative model-enabled decoding process of semantic [PITH_FULL_IMAGE:figures/full_fig_p017_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Three critical domain-specific challenges in generative semantic communications. Challenge 1: Strong channel noise causes unstable [PITH_FULL_IMAGE:figures/full_fig_p018_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Three typical scenarios of diffusion-based semantic communications. (a) Fidelity-oriented human semantic communications aiming to [PITH_FULL_IMAGE:figures/full_fig_p020_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Primary quantitative and qualitative results reproduced from [PITH_FULL_IMAGE:figures/full_fig_p023_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Coordination protocols with diffusion models in intent [PITH_FULL_IMAGE:figures/full_fig_p025_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Promising future research directions for diffusion model-based generative semantic communications. Six key directions span from [PITH_FULL_IMAGE:figures/full_fig_p026_14.png] view at source ↗
read the original abstract

Semantic communications mark a paradigm shift from bit-accurate transmission toward meaning-centric communication, essential as wireless systems approach theoretical capacity limits. The emergence of generative AI has catalyzed generative semantic communications, where receivers reconstruct content from minimal semantic cues by leveraging learned priors. Among generative approaches, diffusion models stand out for their superior generation quality, stable training dynamics, and rigorous theoretical foundations. However, the field currently lacks systematic guidance connecting diffusion techniques to communication system design, forcing researchers to navigate disparate literatures. This article provides the first comprehensive tutorial on diffusion models for generative semantic communications. We present score-based diffusion foundations and systematically review three technical pillars: conditional diffusion for controllable generation, efficient diffusion for accelerated inference, and generalized diffusion for cross-domain adaptation. In addition, we introduce an inverse problem perspective that reformulates semantic decoding as posterior inference, bridging semantic communications with computational imaging. Through analysis of human-centric, machine-centric, and agent-centric scenarios, we illustrate how diffusion models enable extreme compression while maintaining semantic fidelity and robustness. By bridging generative AI innovations with communication system design, this article aims to establish diffusion models as foundational components of next-generation wireless networks and beyond.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. This tutorial manuscript provides a systematic review of diffusion models for generative semantic communications in 6G and beyond. It covers score-based diffusion foundations, reviews three technical pillars (conditional diffusion for controllable generation, efficient diffusion for accelerated inference, and generalized diffusion for cross-domain adaptation), introduces an inverse-problem reformulation of semantic decoding as posterior inference, and analyzes applications across human-centric, machine-centric, and agent-centric scenarios to illustrate extreme compression while preserving semantic fidelity and robustness.

Significance. If the synthesis holds, the paper offers valuable guidance by bridging generative AI and communication system design literatures, with the inverse-problem perspective providing a useful conceptual link to computational imaging. As a tutorial without new empirical results or parameter-free derivations, its primary contribution is organizational rather than foundational; credit is due for the structured coverage of the three pillars and scenario analysis.

major comments (2)
  1. [conditional diffusion pillar] Section on conditional diffusion (pillar 1): the review relies on standard conditioning mechanisms and score-matching derivations from general generative modeling; it does not derive or cite channel-aware likelihood models that replace the forward diffusion process with realistic wireless impairments (fading, interference, non-Gaussian noise), which is required for the posterior-inference claim to remain accurate under varying SNR and mobility conditions.
  2. [inverse problem perspective] Inverse-problem reformulation section: while semantic decoding is recast as posterior inference, the manuscript presents only standard techniques and existing citations without demonstrating or referencing adaptations that preserve semantic fidelity when the diffusion noise is supplanted by wireless channel statistics, leaving the central generalization unvalidated.
minor comments (2)
  1. [efficient diffusion pillar] The efficient diffusion pillar would benefit from quantitative tables comparing sampling steps or latency against communication-relevant metrics such as end-to-end delay under different channel conditions.
  2. [generalized diffusion pillar] Notation for the generalized diffusion pillar could be clarified to explicitly distinguish domain-adaptation losses from standard diffusion objectives.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We are grateful to the referee for the thorough review and valuable suggestions. The feedback helps us improve the manuscript by strengthening the discussion on realistic wireless conditions. We provide detailed responses to the major comments and indicate the revisions we will implement.

read point-by-point responses

  1. Referee: [conditional diffusion pillar] Section on conditional diffusion (pillar 1): the review relies on standard conditioning mechanisms and score-matching derivations from general generative modeling; it does not derive or cite channel-aware likelihood models that replace the forward diffusion process with realistic wireless impairments (fading, interference, non-Gaussian noise), which is required for the posterior-inference claim to remain accurate under varying SNR and mobility conditions.

    Authors: We appreciate this observation. As a tutorial paper, our aim is to synthesize and organize existing techniques rather than derive new models. The section on conditional diffusion reviews standard mechanisms and their use in semantic communications. We acknowledge the importance of channel-aware adaptations for accurate modeling under wireless impairments. In the revised version, we will expand this section to include a discussion of channel-aware likelihood models, citing relevant literature on diffusion models adapted for wireless channels and semantic communications under fading and noise conditions. This will better justify the posterior inference approach in practical scenarios. revision: yes

  2. Referee: [inverse problem perspective] Inverse-problem reformulation section: while semantic decoding is recast as posterior inference, the manuscript presents only standard techniques and existing citations without demonstrating or referencing adaptations that preserve semantic fidelity when the diffusion noise is supplanted by wireless channel statistics, leaving the central generalization unvalidated.

    Authors: Thank you for highlighting this point. The inverse-problem perspective is introduced to bridge semantic communications with computational imaging by reformulating decoding as posterior inference. While we present standard techniques, we agree that more explicit references to adaptations for wireless channel statistics are needed to maintain semantic fidelity. We will revise this section to reference and discuss existing works that adapt diffusion models to replace Gaussian diffusion noise with channel-induced distortions, including examples from literature on robust semantic decoding. This will provide better support for the generalization without requiring new empirical validation, consistent with the tutorial nature of the manuscript. revision: yes

Circularity Check

0 steps flagged

Tutorial review of diffusion models for semantic communications is self-contained with no circular derivations

full rationale

The manuscript is structured as a tutorial that presents score-based diffusion foundations drawn from prior literature and systematically reviews three established technical pillars (conditional diffusion, efficient diffusion, generalized diffusion) plus an inverse-problem reformulation of semantic decoding. All load-bearing technical content is attributed to external references rather than derived from parameters fitted inside this paper or from self-referential equations that rename inputs as outputs. No self-definitional steps, fitted-input predictions, or uniqueness theorems imported solely via author self-citation appear in the derivation chain. The work therefore remains independent of its own fitted values and is self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

This is a review/tutorial paper that draws on established literature in generative AI and communications without introducing new free parameters, axioms, or invented entities of its own.

pith-pipeline@v0.9.0 · 5536 in / 1059 out tokens · 42923 ms · 2026-05-17T23:36:55.283304+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

179 extracted references · 179 canonical work pages · 4 internal anchors

  1. [1]

    Recent contributions to the mathematical theory of communication,

    W. Weaver, “Recent contributions to the mathematical theory of communication,”ETC: A Review of General Semantics, vol. 10, no. 4, pp. 261–281, 1953

    work page 1953

  2. [2]

    A survey on semantic communication networks: Architecture, security, and privacy,

    S. Guo, Y . Wang, N. Zhang, Z. Su, T. H. Luan, Z. Tian, and X. Shen, “A survey on semantic communication networks: Architecture, security, and privacy,”IEEE Commun. Surv. Tut., vol. 27, no. 5, pp. 2860–2894, 2024

    work page 2024

  3. [3]

    A contemporary survey on semantic communications: Theory of mind, generative AI, and deep joint source-channel coding,

    L. X. Nguyen, A. D. Raha, P. S. Aung, D. Niyato, Z. Han, and C. S. Hong, “A contemporary survey on semantic communications: Theory of mind, generative AI, and deep joint source-channel coding,”IEEE Commun. Surv. Tut., 2025, Early Access

    work page 2025

  4. [4]

    Intellicise wireless networks from semantic communications: A survey, research issues, and challenges,

    P. Zhang, W. Xu, Y . Liu, X. Qin, K. Niu, S. Cui, G. Shi, Z. Qin, X. Xu, F. Wang, et al., “Intellicise wireless networks from semantic communications: A survey, research issues, and challenges,”IEEE Commun. Surv. Tut., vol. 27, no. 3, pp. 2051–2084, 2024

    work page 2051

  5. [5]

    Less data, more knowledge: Building next-generation semantic communication networks,

    C. Chaccour, W. Saad, M. Debbah, Z. Han, and H. V . Poor, “Less data, more knowledge: Building next-generation semantic communication networks,”IEEE Commun. Surv. Tut., vol. 27, no. 1, pp. 37–76, 2024

    work page 2024

  6. [6]

    Informed machine learning–a taxonomy and survey of integrating prior knowledge into learning systems,

    L. von Rueden, S. Mayer, K. Beckh, B. Georgiev, S. Giesselbach, R. Heese, B. Kirsch, J. Pfrommer, A. Pick, R. Ramamurthy, et al., “Informed machine learning–a taxonomy and survey of integrating prior knowledge into learning systems,”IEEE Trans. Knowl. Data Eng., vol. 35, no. 1, pp. 614–633, 2021

    work page 2021

  7. [7]

    Generative AI meets semantic communication: Evolution and revolution of communication tasks,

    E. Grassucci, J. Park, S. Barbarossa, S. L. Kim, J. Choi, and D. Com- miniello, “Generative AI meets semantic communication: Evolution and revolution of communication tasks,”arXiv preprint arXiv:2401.06803, 2024

    work page arXiv 2024

  8. [8]

    Generative semantic communication: Architec- tures, technologies, and applications,

    J. Ren, Y . Sun, H. Du, W. Yuan, C. Wang, X. Wang, Y . Zhou, Z. Zhu, F. Wang, and S. Cui, “Generative semantic communication: Architec- tures, technologies, and applications,”arXiv preprint arXiv:2412.08642, 2024

    work page arXiv 2024

  9. [9]

    Deep generative modeling reshapes compression and transmission: From efficiency to resiliency,

    J. Dai, X. Qin, S. Wang, L. Xu, K. Niu, and P. Zhang, “Deep generative modeling reshapes compression and transmission: From efficiency to resiliency,”IEEE Wireless Commun., vol. 31, no. 4, pp. 48–56, 2024

    work page 2024

  10. [10]

    Improving language understanding by generative pre-training,

    A. Radford, K. Narasimhan, T. Salimans, and I. Sutskever, “Improving language understanding by generative pre-training,” Tech. Rep., OpenAI, 2018

    work page 2018

  11. [11]

    Language models are unsupervised multitask learners,

    A. Radford, J. Wu, R. Child, D. Luan, D. Amodei, and I. Sutskever, “Language models are unsupervised multitask learners,”OpenAI Blog, vol. 1, no. 8, pp. 9, 2019

    work page 2019

  12. [12]

    Language models are few-shot learners,

    T. Brown, B. Mann, N. Ryder, M. Subbiah, J. D. Kaplan, P. Dhariwal, A. Neelakantan, P. Shyam, G. Sastry, A. Askell, et al., “Language models are few-shot learners,” inProc. Int. Conf. Neural Inf. Process. Syst. (NeurIPS), 2020, pp. 1877–1901

    work page 2020

  13. [13]

    Deep unsupervised learning using nonequilibrium thermodynamics,

    J. Sohl-Dickstein, E. Weiss, N. Maheswaranathan, and S. Ganguli, “Deep unsupervised learning using nonequilibrium thermodynamics,” in Proc. Int. Conf. Mach. Learn. (ICML). PMLR, 2015, pp. 2256–2265

    work page 2015

  14. [14]

    Diffusion-aided joint source channel coding for high realism wireless image transmission,

    M. Yang, B. Liu, B. Wang, and H. S. Kim, “Diffusion-aided joint source channel coding for high realism wireless image transmission,” arXiv preprint arXiv:2404.17736, 2024

    work page arXiv 2024

  15. [15]

    CDDM: Channel denoising diffusion models for wireless semantic communications,

    T. Wu, Z. Chen, D. He, L. Qian, Y . Xu, M. Tao, and W. Zhang, “CDDM: Channel denoising diffusion models for wireless semantic communications,”IEEE Trans. Wireless Commun., vol. 23, no. 9, pp. 11168–11183, 2024

    work page 2024

  16. [16]

    Diffusion-driven semantic communication for generative models with bandwidth constraints,

    L. Guo, W. Chen, Y . Sun, B. Ai, N. Pappas, and T. Q. S. Quek, “Diffusion-driven semantic communication for generative models with bandwidth constraints,”IEEE Trans. Wireless Commun., vol. 24, no. 8, pp. 6490–6503, 2025

    work page 2025

  17. [17]

    Enhancing deep reinforcement learning: A tutorial on generative diffusion models in network optimization,

    H. Du, R. Zhang, Y . Liu, J. Wang, Y . Lin, Z. Li, D. Niyato, J. Kang, Z. Xiong, S. Cui, et al., “Enhancing deep reinforcement learning: A tutorial on generative diffusion models in network optimization,”IEEE Commun. Surv. Tut., vol. 26, no. 4, pp. 2611–2646, 2024

    work page 2024

  18. [18]

    Generative diffusion models for wireless networks: Fundamental, architecture, and state-of-the-art,

    D. Fan, R. Meng, X. Xu, Y . Liu, G. Nan, C. Feng, S. Han, S. Gao, B. Xu, D. Niyato, et al., “Generative diffusion models for wireless networks: Fundamental, architecture, and state-of-the-art,”arXiv preprint arXiv:2507.16733, 2025

    work page arXiv 2025

  19. [19]

    DiffSG: A generative solver for network optimization with diffusion model,

    R. Liang, B. Yang, Z. Yu, B. Guo, X. Cao, M. Debbah, H. V . Poor, and C. Yuen, “DiffSG: A generative solver for network optimization with diffusion model,”IEEE Commun. Mag., vol. 63, no. 6, pp. 16–24, 2025

    work page 2025

  20. [20]

    An introduction to variational autoencoders,

    D. P. Kingma and M. Welling, “An introduction to variational autoencoders,”Found. Trends Mach. Learn., vol. 12, no. 4, pp. 307–392, 2019

    work page 2019

  21. [21]

    Estimation of non-normalized statistical models by score matching,

    A. Hyv ¨arinen, “Estimation of non-normalized statistical models by score matching,”J. Mach. Learn. Res., vol. 6, no. 24, pp. 695–709, 2005

    work page 2005

  22. [22]

    A tutorial on energy-based learning,

    Y . LeCun, S. Chopra, R. Hadsell, M. Ranzato, and F. Huang, “A tutorial on energy-based learning,” inPredicting Structured Data, G. Bakir, T. Hofmann, B. Sch ¨olkopf, A. Smola, and B. Taskar, Eds. MIT Press, 2006

    work page 2006

  23. [23]

    A learning algorithm for Boltzmann machines,

    D. H. Ackley, G. E. Hinton, and T. J. Sejnowski, “A learning algorithm for Boltzmann machines,”Cogn. Sci., vol. 9, no. 1, pp. 147–169, 1985

    work page 1985

  24. [24]

    A simple introduction to Markov Chain Monte Carlo sampling,

    D. van Ravenzwaaij, P. Cassey, and S. D. Brown, “A simple introduction to Markov Chain Monte Carlo sampling,”Psychon. Bull. Rev., vol. 25, no. 1, pp. 143–154, 2018

    work page 2018

  25. [25]

    Annealed importance sampling,

    R. M. Neal, “Annealed importance sampling,”Stat. Comput., vol. 11, no. 2, pp. 125–139, 2001

    work page 2001

  26. [26]

    Autoregressive models in vision: A survey,

    J. Xiong, G. Liu, L. Huang, C. Wu, T. Wu, Y . Mu, Y . Yao, H. Shen, Z. Wan, J. Huang, et al., “Autoregressive models in vision: A survey,” arXiv preprint arXiv:2411.05902, 2024

    work page arXiv 2024

  27. [27]

    Auto-encoding variational Bayes,

    D. P. Kingma and M. Welling, “Auto-encoding variational Bayes,” in Proc. Int. Conf. Learn. Represent. (ICLR), 2014

    work page 2014

  28. [28]

    Pixel recurrent neural networks,

    A. van den Oord, N. Kalchbrenner, and K. Kavukcuoglu, “Pixel recurrent neural networks,” inProc. Int. Conf. Mach. Learn. (ICML). PMLR, 2016, pp. 1747–1756

    work page 2016

  29. [29]

    Generative adversarial nets,

    I. Goodfellow, J. Pouget-Abadie, M. Mirza, B. Xu, D. Warde-Farley, S. Ozair, A. Courville, and Y . Bengio, “Generative adversarial nets,” inProc. Int. Conf. Neural Inf. Process. Syst. (NeurIPS), 2014

    work page 2014

  30. [30]

    Improved techniques for training GANs,

    T. Salimans, I. Goodfellow, W. Zaremba, V . Cheung, A. Radford, and X. Chen, “Improved techniques for training GANs,” inProc. Int. Conf. Neural Inf. Process. Syst. (NeurIPS), 2016

    work page 2016

  31. [31]

    A bound for the error in the normal approximation to the distribution of a sum of dependent random variables,

    C. Stein, “A bound for the error in the normal approximation to the distribution of a sum of dependent random variables,” inProc. Sixth Berkeley Symp. Math. Statist. Probab.University of California Press, 1972, vol. 6, pp. 583–603

    work page 1972

  32. [32]

    How to train your energy-based models,

    Y . Song and D. P. Kingma, “How to train your energy-based models,” arXiv preprint arXiv:2101.03288, 2021

    work page arXiv 2021

  33. [33]

    The Principles of Diffusion Models

    C. H. Lai, Y . Song, D. Kim, Y . Mitsufuji, and S. Ermon, “The principles of diffusion models,”arXiv preprint arXiv:2510.21890, 2025

    work page internal anchor Pith review arXiv 2025

  34. [34]

    A connection between score matching and denoising autoencoders,

    P. Vincent, “A connection between score matching and denoising autoencoders,”Neural Comput., vol. 23, no. 7, pp. 1661–1674, 2011

    work page 2011

  35. [35]

    Bayesian learning via stochastic gradient Langevin dynamics,

    M. Welling and Y . W. Teh, “Bayesian learning via stochastic gradient Langevin dynamics,” inProc. Int. Conf. Mach. Learn. (ICML). PMLR, 2011, pp. 681–688

    work page 2011

  36. [36]

    Score-based generative modeling through stochastic differential equations,

    Y . Song, J. Sohl-Dickstein, D. P. Kingma, A. Kumar, S. Ermon, and B. Poole, “Score-based generative modeling through stochastic differential equations,” inProc. Int. Conf. Learn. Represent. (ICLR), 2021

    work page 2021

  37. [37]

    Reverse-time diffusion equation models,

    B. D. O. Anderson, “Reverse-time diffusion equation models,”Stochas- tic Process. Appl., vol. 12, no. 3, pp. 313–326, 1982. 28

    work page 1982

  38. [38]

    Reasons for the superiority of stochastic estimators over deterministic ones: Robustness, consistency and perceptual quality,

    G. Ohayon, T. J. Adrai, M. Elad, and T. Michaeli, “Reasons for the superiority of stochastic estimators over deterministic ones: Robustness, consistency and perceptual quality,” inProc. Int. Conf. Mach. Learn. (ICML). PMLR, 2023, pp. 26474–26494

    work page 2023

  39. [39]

    Generative modeling by estimating gradients of the data distribution,

    Y . Song and S. Ermon, “Generative modeling by estimating gradients of the data distribution,” inProc. Int. Conf. Neural Inf. Process. Syst. (NeurIPS), 2019

    work page 2019

  40. [40]

    Denoising diffusion probabilistic models,

    J. Ho, A. Jain, and P. Abbeel, “Denoising diffusion probabilistic models,” inProc. Int. Conf. Neural Inf. Process. Syst. (NeurIPS), 2020, pp. 6840–6851

    work page 2020

  41. [41]

    Neural ordinary differential equations,

    R. T. Q. Chen, Y . Rubanova, J. Bettencourt, and D. K. Duvenaud, “Neural ordinary differential equations,” inProc. Int. Conf. Neural Inf. Process. Syst. (NeurIPS), 2018

    work page 2018

  42. [42]

    Diffusion models beat GANs on image synthesis,

    P. Dhariwal and A. Nichol, “Diffusion models beat GANs on image synthesis,” inProc. Int. Conf. Neural Inf. Process. Syst. (NeurIPS), 2021, pp. 8780–8794

    work page 2021

  43. [43]

    ILVR: Condi- tioning method for denoising diffusion probabilistic models,

    J. Choi, S. Kim, Y . Jeong, Y . Gwon, and S. Yoon, “ILVR: Condi- tioning method for denoising diffusion probabilistic models,” inProc. IEEE/CVF Int. Conf. Comput. Vis. (ICCV), 2021, pp. 14367–14376

    work page 2021

  44. [44]

    SDEdit: Guided image synthesis and editing with stochastic differential equations,

    C. Meng, Y . He, Y . Song, J. Song, J. Wu, J. Y . Zhu, and S. Ermon, “SDEdit: Guided image synthesis and editing with stochastic differential equations,” inProc. Int. Conf. Learn. Represent. (ICLR), 2022

    work page 2022

  45. [45]

    RePaint: Inpainting using denoising diffusion probabilistic models,

    A. Lugmayr, M. Danelljan, A. Romero, F. Yu, R. Timofte, and L. Van Gool, “RePaint: Inpainting using denoising diffusion probabilistic models,” inProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), 2022, pp. 11461–11471

    work page 2022

  46. [46]

    Denoising diffusion restoration models,

    B. Kawar, M. Elad, S. Ermon, and J. Song, “Denoising diffusion restoration models,” inProc. Int. Conf. Neural Inf. Process. Syst. (NeurIPS), 2022, pp. 23593–23606

    work page 2022

  47. [47]

    Improving diffusion models for inverse problems using manifold constraints,

    H. Chung, B. Sim, D. Ryu, and J. C. Ye, “Improving diffusion models for inverse problems using manifold constraints,” inProc. Int. Conf. Neural Inf. Process. Syst. (NeurIPS), 2022, pp. 25683–25696

    work page 2022

  48. [48]

    FreeDoM: Training- free energy-guided conditional diffusion model,

    J. Yu, Y . Wang, C. Zhao, B. Ghanem, and J. Zhang, “FreeDoM: Training- free energy-guided conditional diffusion model,” inProc. IEEE/CVF Int. Conf. Comput. Vis. (ICCV), 2023, pp. 23174–23184

    work page 2023

  49. [49]

    Refining generative process with discriminator guidance in score-based diffusion models,

    D. Kim, Y . Kim, S. J. Kwon, W. Kang, and I. C. Moon, “Refining generative process with discriminator guidance in score-based diffusion models,” inProc. Int. Conf. Mach. Learn. (ICML), 2023, vol. 202, pp. 16567–16598

    work page 2023

  50. [50]

    Diffusion posterior sampling for general noisy inverse problems,

    H. Chung, J. Kim, M. T. Mccann, M. L. Klasky, and J. C. Ye, “Diffusion posterior sampling for general noisy inverse problems,” inProc. Int. Conf. Learn. Represent. (ICLR), 2023

    work page 2023

  51. [51]

    Pseudoinverse-guided diffusion models for inverse problems,

    J. Song, A. Vahdat, M. Mardani, and J. Kautz, “Pseudoinverse-guided diffusion models for inverse problems,” inProc. Int. Conf. Learn. Represent. (ICLR), 2023

    work page 2023

  52. [52]

    Solving linear inverse problems provably via posterior sampling with latent diffusion models,

    L. Rout, N. Raoof, G. Daras, C. Caramanis, A. Dimakis, and S. Shakkot- tai, “Solving linear inverse problems provably via posterior sampling with latent diffusion models,” inProc. Int. Conf. Neural Inf. Process. Syst. (NeurIPS), 2024

    work page 2024

  53. [53]

    A variational perspective on solving inverse problems with diffusion models,

    M. Mardani, J. Song, J. Kautz, and A. Vahdat, “A variational perspective on solving inverse problems with diffusion models,” inProc. Int. Conf. Learn. Represent. (ICLR), 2024

    work page 2024

  54. [54]

    Deep equilibrium diffusion restoration with parallel sampling,

    J. Cao, Y . Shi, K. Zhang, Y . Zhang, R. Timofte, and L. Van Gool, “Deep equilibrium diffusion restoration with parallel sampling,” inProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), 2024, pp. 2824–2834

    work page 2024

  55. [55]

    Improving diffusion inverse problem solving with decoupled noise annealing,

    B. Zhang, W. Chu, J. Berner, C. Meng, A. Anandkumar, and Y . Song, “Improving diffusion inverse problem solving with decoupled noise annealing,” inProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), 2025, pp. 20895–20905

    work page 2025

  56. [56]

    SitCom: Step-wise triple-consistent diffusion sampling for inverse problems,

    I. Alkhouri, S. Liang, C. H. Huang, J. Dai, Q. Qu, S. Ravishankar, and R. Wang, “SitCom: Step-wise triple-consistent diffusion sampling for inverse problems,” inProc. Int. Conf. Mach. Learn. (ICML), 2025

    work page 2025

  57. [57]

    Classifier-free diffusion guidance,

    J. Ho and T. Salimans, “Classifier-free diffusion guidance,” inProc. Int. Conf. Neural Inf. Process. Syst. (NeurIPS), 2021

    work page 2021

  58. [58]

    High- resolution image synthesis with latent diffusion models,

    R. Rombach, A. Blattmann, D. Lorenz, P. Esser, and B. Ommer, “High- resolution image synthesis with latent diffusion models,” inProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), 2022, pp. 10684–10695

    work page 2022

  59. [59]

    GLIGEN: Open-set grounded text-to-image generation,

    Y . Li, H. Liu, Q. Wu, F. Mu, J. Yang, J. Gao, C. Li, and Y . J. Lee, “GLIGEN: Open-set grounded text-to-image generation,” in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), 2023, pp. 22511–22521

    work page 2023

  60. [60]

    InstructPix2Pix: Learning to follow image editing instructions,

    T. Brooks, A. Holynski, and A. A. Efros, “InstructPix2Pix: Learning to follow image editing instructions,” inProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), 2023, pp. 18392–18402

    work page 2023

  61. [61]

    Shap-E: Generating Conditional 3D Implicit Functions

    H. Jun and A. Nichol, “Shap-E: Generating conditional 3D implicit functions,”arXiv preprint arXiv:2305.02463, 2023

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  62. [62]

    Scalable diffusion models with transformers,

    W. Peebles and S. Xie, “Scalable diffusion models with transformers,” in Proc. IEEE/CVF Int. Conf. Comput. Vis. (ICCV), 2023, pp. 4195–4205

    work page 2023

  63. [63]

    MDTv2: Masked diffusion transformer is a strong image synthesizer,

    S. Gao, P. Zhou, M. M. Cheng, and S. Yan, “MDTv2: Masked diffusion transformer is a strong image synthesizer,” inProc. IEEE Int. Conf. Comput. Vis. (ICCV), 2023

    work page 2023

  64. [64]

    Adding conditional control to text-to-image diffusion models,

    L. Zhang, A. Rao, and M. Agrawala, “Adding conditional control to text-to-image diffusion models,” inProc. IEEE/CVF Int. Conf. Comput. Vis. (ICCV), 2023, pp. 3836–3847

    work page 2023

  65. [65]

    T2I- Adapter: Learning adapters to dig out more controllable ability for text-to-image diffusion models,

    C. Mou, X. Wang, L. Xie, Y . Wu, J. Zhang, Z. Qi, and Y . Shan, “T2I- Adapter: Learning adapters to dig out more controllable ability for text-to-image diffusion models,” inProc. AAAI Conf. Artif. Intell., 2024, vol. 38, pp. 4296–4304

    work page 2024

  66. [66]

    AnimateDiff: Animate your personalized text-to- image diffusion models without specific tuning,

    Y . Guo, C. Yang, A. Rao, Z. Liang, Y . Wang, Y . Qiao, M. Agrawala, D. Lin, and B. Dai, “AnimateDiff: Animate your personalized text-to- image diffusion models without specific tuning,” inProc. Int. Conf. Learn. Represent. (ICLR), 2024

    work page 2024

  67. [67]

    LLM-grounded video diffusion models,

    L. Lian, B. Shi, A. Yala, T. Darrell, and B. Li, “LLM-grounded video diffusion models,” inProc. Int. Conf. Learn. Represent. (ICLR), 2024

    work page 2024

  68. [68]

    SEINE: Short-to-long video diffusion model for generative transition and prediction,

    X. Chen, Y . Wang, L. Zhang, S. Zhuang, X. Ma, J. Yu, Y . Wang, D. Lin, Y . Qiao, and Z. Liu, “SEINE: Short-to-long video diffusion model for generative transition and prediction,” inProc. Int. Conf. Learn. Represent. (ICLR), 2024

    work page 2024

  69. [69]

    PixArt- σ: Weak-to-strong training of diffusion transformer for 4K text-to-image generation,

    J. Chen, C. Ge, E. Xie, Y . Wu, L. Yao, X. Ren, Z. Wang, P. Luo, H. Lu, and Z. Li, “PixArt- σ: Weak-to-strong training of diffusion transformer for 4K text-to-image generation,” inProc. Eur. Conf. Comput. Vis. (ECCV). Springer, 2024, pp. 74–91

    work page 2024

  70. [70]

    Direct discriminative optimization: Your likelihood-based visual generative model is secretly a GAN discriminator,

    K. Zheng, Y . Chen, H. Chen, G. He, M. Y . Liu, J. Zhu, and Q. Zhang, “Direct discriminative optimization: Your likelihood-based visual generative model is secretly a GAN discriminator,” inProc. Int. Conf. Mach. Learn. (ICML), 2025

    work page 2025

  71. [71]

    T2V-Turbo-v2: Enhancing video generation model post-training through data, reward, and conditional guidance design,

    J. Li, Q. Long, J. Zheng, X. Gao, R. Piramuthu, W. Chen, and W. Y . Wang, “T2V-Turbo-v2: Enhancing video generation model post-training through data, reward, and conditional guidance design,” inProc. Int. Conf. Learn. Represent. (ICLR), 2025

    work page 2025

  72. [72]

    Denoising diffusion implicit models,

    J. Song, C. Meng, and S. Ermon, “Denoising diffusion implicit models,” inProc. Int. Conf. Learn. Represent. (ICLR), 2021

    work page 2021

  73. [73]

    Sparse MRI: The application of compressed sensing for rapid MR imaging,

    M. Lustig, D. Donoho, and J. M. Pauly, “Sparse MRI: The application of compressed sensing for rapid MR imaging,”Magn. Reson. Med., vol. 58, no. 6, pp. 1182–1195, 2007

    work page 2007

  74. [74]

    Tweedie’s formula and selection bias,

    B. Efron, “Tweedie’s formula and selection bias,”J. Amer. Statist. Assoc., vol. 106, no. 496, pp. 1602–1614, 2011

    work page 2011

  75. [75]

    Parallel diffusion models of operator and image for blind inverse problems,

    H. Chung, J. Kim, S. Kim, and J. C. Ye, “Parallel diffusion models of operator and image for blind inverse problems,” inProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), 2023, pp. 6059–6069

    work page 2023

  76. [76]

    Reduce, reuse, recycle: Compositional generation with energy-based diffusion models and MCMC,

    Y . Du, C. Durkan, R. Strudel, J. B. Tenenbaum, S. Dieleman, R. Fergus, J. Sohl-Dickstein, A. Doucet, and W. S. Grathwohl, “Reduce, reuse, recycle: Compositional generation with energy-based diffusion models and MCMC,” inProc. Int. Conf. Mach. Learn. (ICML). PMLR, 2023, pp. 8489–8510

    work page 2023

  77. [77]

    A Survey on Diffusion Models for Inverse Problems

    G. Daras, H. Chung, C. H. Lai, Y . Mitsufuji, J. C. Ye, P. Milanfar, A. G. Dimakis, and M. Delbracio, “A survey on diffusion models for inverse problems,”arXiv preprint arXiv:2410.00083, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  78. [78]

    Wavelet score-based generative modeling,

    F. Guth, S. Coste, V . De Bortoli, and S. Mallat, “Wavelet score-based generative modeling,” inProc. Int. Conf. Neural Inf. Process. Syst. (NeurIPS), 2022, pp. 478–491

    work page 2022

  79. [79]

    Wavelet diffusion models are fast and scalable image generators,

    H. Phung, Q. Dao, and A. Tran, “Wavelet diffusion models are fast and scalable image generators,” inProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), 2023, pp. 10199–10208

    work page 2023

  80. [80]

    LMD: Faster image reconstruction with latent masking diffusion,

    Z. Ma, Z. Yu, J. Li, and B. Zhou, “LMD: Faster image reconstruction with latent masking diffusion,” inProc. AAAI Conf. Artif. Intell., 2024, vol. 38, pp. 4145–4153

    work page 2024

Showing first 80 references.