Pith. sign in

REVIEW 3 major objections 5 minor 66 references

Discrete Noise Inversion for Next-scale Autoregressive Text-based Image Editing

T0 review · 3 major / 5 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read VARIN proposes a pseudo-inverse of the Gumbel argmax that recovers source image noise, enabling training-free text-based editing for visual autoregressive models while preserving background details.

desk verdict LAI is a genuinely new pseudo-inverse for argmax with an exact reconstruction guarantee, but the headline comparison uses a mismatched Regeneration baseline; the central idea survives, the evidence as presented does not. read the letter →

arxiv 2509.01984 v2 pith:XM67MBKQ submitted 2025-09-02 cs.CV

Quan Dao , Xiaoxiao He , Ligong Han , Ngan Hoai Nguyen , Amin Heyrani Nobar , Faez Ahmed , Han Zhang , Viet Anh Nguyen
show 1 more author
Dimitris Metaxas
This is my paper
classification cs.CV
keywords visualautoregressivemodelstext-basedimageeditingnoiseinversionGumbel-maxtrickargmaxpseudo-inversetraining-freenext-scalepredictionHART
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 introduces VARIN, a training-free way to edit images with visual autoregressive (VAR) models by inverting the discrete sampling step. Because VAR models generate tokens by argmax over Gumbel-perturbed logits, the argmax cannot be inverted uniquely; the paper proposes Location-aware Argmax Inversion (LAI), a pseudo-inverse that recovers Gumbel noises which reproduce the source token maps exactly while remaining close to standard Gumbel noise. During editing, VARIN starts from the source token maps through scale 6, then mixes recovered inverse noises with fresh Gumbel noise under the target prompt, with the inverse-noise weight decaying to zero by the final scale. On PIE-Bench with HART, this beats the regeneration baseline on structure distance (11.46 vs 25.56) and background PSNR (26.54 vs 20.45) while matching prompt alignment, and runs in about one second per image. The paper argues this makes discrete autoregressive image models practical for prompt-guided editing without fine-tuning or optimization.

What carries the argument

Location-aware Argmax Inversion (LAI): a pseudo-inverse of the argmax sampling step. For each token it draws the winning value from Gumbel(p_label, 1) and every losing value from a Gumbel truncated at winning_value - tau. The truncation enforces the source label as argmax, the shared location keeps the noise Gumbel-like, and tau is the dial that trades edit freedom against background preservation.

What would settle it

Take a held-out set of source images and target prompts with large structural changes (e.g., pose or object interaction), run VARIN with the paper's fixed settings (s=6, tau=18), and measure whether CLIP edit similarity and background PSNR remain near PIE-Bench levels; the paper's own failure cases show such prompts degrade, so a systematic curve would bound the claim.

Watch

Extended reading notes

Core claim

VARIN's central claim is that a discrete next-scale autoregressive model can be edited by noise inversion if one replaces the non-invertible argmax of the Gumbel-max trick with a location-aware pseudo-inverse. LAI samples the perturbed logit at the ground-truth token from a Gumbel centered at the model's predicted logit, and samples all other positions from a Gumbel truncated so they cannot exceed that value minus tau. This guarantees argmax equals the source token, so the extracted noises reconstruct the source image token maps perfectly; because the sampling locations come from the model's own logits, the noises behave like standard Gumbel noise, and tau controls how much source bias they

Load-bearing premise

The load-bearing premise is that interpolating the reconstructed inverse Gumbel noises with fresh Gumbel noise, under a fixed start scale and decay schedule, produces semantically meaningful target-aligned edits, and that this behavior holds across VAR architectures and prompts; the paper validates this only empirically on PIE-Bench.

Editorial extensions

If this is right

  • VARIN gives any next-scale VAR model (demonstrated with HART and Switti) a training-free text-editing capability, so editing costs no extra optimization or fine-tuning.
  • Because inverse noises reconstruct source token maps exactly, edits can be localized to scales 6 and above while early-scale layout tokens remain untouched.
  • The tau parameter directly controls the preservation-editability trade-off, with tau around 14-20 giving the best balance on PIE-Bench.
  • Editing runs in roughly one second per image, about 10x faster than DDPM-Inversion, making real-time interactive editing plausible.
  • The approach is orthogonal to attention-control and test-time optimization methods and can be combined with them for stronger edits.

Reading between the lines

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

  • If the mechanism transfers as claimed, the same LAI pseudo-inverse should also work for next-token (not just next-scale) discrete autoregressive image models, since the Gumbel-max inversion is scale-agnostic.
  • The linear lambda schedule and the fixed start scale 6 are chosen empirically from PIE-Bench; prompts or images with very different structure (e.g., large pose changes) may need per-input tau or a later start scale to avoid failures.
  • A testable extension is to make tau image-adaptive (e.g., set by estimated edit-region size) rather than a global constant, which could improve preservation on sparse edits.
  • The method's comparison to optimization-based diffusion editing suggests that discrete inversion alone can reach similar fidelity, but head-to-head user studies on harder structural edits would clarify where the ceiling is.
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 VARIN, a training-free editing method for next-scale visual autoregressive models such as HART and Switti. The core idea is to invert the Gumbel-max sampling procedure used by discrete VAR models, via a new Location-aware Argmax Inversion (LAI) that produces inverse noises able to reconstruct the source token maps. Editing is performed by regenerating scales from a start scale s while interpolating fresh Gumbel noise with the inverse noise, with a hyperparameter τ controlling how much source information is retained. The method is evaluated on PIE-Bench against Regeneration, DICE, EditAR, and several diffusion-based editing baselines, reporting better background preservation and competitive prompt alignment, with a user study and qualitative examples. The reconstruction guarantee of LAI is mathematically sound by construction, and the method is simple and fast, but the quantitative evaluation contains an internal inconsistency in the Regeneration baseline and some claims are not supported by the reported numbers.

Significance. If the claims hold, this is a useful contribution: it is, to my knowledge, the first training-free noise-inversion editing method designed for next-scale autoregressive visual models, and it demonstrates that a simple pseudo-inverse construction can yield reasonable edits with real-time speed. The reconstruction property of LAI is correct by construction, and the paper shows qualitative generalization to a second VAR architecture (Switti), which is a strength. The main weakness is that the headline quantitative claims are weakened by a mismatched baseline and by hyperparameters selected on the same evaluation set used for the final table. The underlying method is plausible, and the reconstruction guarantee is rigorous, but the evaluation needs to be redone or substantially qualified before the claims can be accepted.

major comments (3)
  1. [Section 5.2, Table 1, Appendix Table 3] The Regeneration row in Table 1 (Structure 25.56, PSNR 20.45, LPIPS 106.50, CLIP Edited 21.13) is exactly the s=8 row of Table 3, not the stated default s=6. At s=6, Regeneration has CLIP Edited 23.11 and Whole 26.83, while VARIN reports 21.49 and 25.05. Therefore the text's claim that VARIN 'adher[es] to the target prompt more effectively' than Regeneration is not supported at the stated default setting. The background-preservation comparison at matched s=6 still favors VARIN (PSNR 26.54 vs 19.21), but the headline claim of simultaneous alignment and preservation must be re-evaluated with matched baselines.
  2. [Section 5.2, Table 1] The same claim of 'adhering to the target prompt more effectively' is also contradicted by the DDPM-Inversion row: DDPM-Inversion has CLIP Edited 23.02 and CLIP Whole 26.22, both higher than VARIN's 21.49 and 25.05. The sentence in Section 5.2 asserting that VARIN 'adher[es] to the target prompt more effectively' than DDPM-Inversion is not supported by the table. Please qualify the claim or present the comparison on the quantities that are actually favorable.
  3. [Section 5.2, Appendix C.2, Table 4] The editing hyperparameters s=6 and τ=18 are selected based on ablations on PIE-Bench, and the final quantitative results are reported on the same PIE-Bench set, with no validation split. In particular, Appendix C.2 explicitly chooses τ based on 'best visual editing results' in Table 4, and Section 5.2 sets s=6 based on qualitative observation on the same benchmark. This test-set tuning risks optimistic bias in the reported margins. I ask the authors to provide a held-out validation protocol or to discuss the sensitivity of the claimed improvements to the chosen hyperparameters.
minor comments (5)
  1. [Algorithm 2] The pseudocode for VARIN inversion does not list csrc as input and does not define p_t. In the main text and Figure 3 it is clear that p_t is computed from the source prompt, but the algorithm should explicitly include 'Input: csrc' and a line such as 'pt ← pθ(·|r<t, csrc)' for reproducibility.
  2. [Algorithm 6] Algorithm 6 uses 'Itgt' in line 1 where the source image must be intended (the input list has Isrc), and the label 'Regeneration:' above the algorithm is misleading. Please correct.
  3. [Various] Typos: 'staring scale' in Algorithm 3, 'editings = 6' in Section 5.2, 'unormalized' and 'algins' in Section 3.2/4.2. These should be fixed.
  4. [Table 6] The user study reports percentages from 25 people on 10 images without confidence intervals or significance testing. Please add uncertainty estimates or at least state the small-sample limitation.
  5. [Section 4.1] The phrase 'Parallel Autoregressive Inversion' is confusing because the loop in Algorithm 2 is sequential over scales. Please clarify what 'parallel' refers to.

Circularity Check

1 steps flagged · score 4.0 of 10

Test-set hyperparameter selection is the only circularity-like step: τ=18 (and s=6) are chosen by ablations on PIE-Bench, and Table 1 reports the selected configuration as the headline result. LAI itself is constructive, and self-citations to DICE are not load-bearing. A separate, non-circular comparison mismatch (Regeneration row equals s=8, not the stated default s=6) further weakens the prompt-

  1. fitted input called prediction [Section 5.2, Appendix C.2, Tables 4 and 1]
    "For both Regeneration and V ARIN, by default, we set the start scale of editings = 6based on observation from Figure 2and set the τ to be 18 in LAI. ... Since the hyperparameter τ is crucial for editing, we perform an ablation study on it in Table 4."

    Table 4's τ=18 row reports Structure 11.46, PSNR 26.54, LPIPS 54.04, MSE 38.33, SSIM 85.39, CLIP Edited 21.49; Table 1's VARIN row reports exactly the same numbers. The ablation is performed on the same PIE-Bench benchmark used for the final comparison, and τ=18 is selected from that sweep as the best trade-off. Consequently, the headline 'VARIN outperforms Regeneration' is not an out-of-sample prediction of a fixed method; it is the in-sample value of a hyperparameter chosen on the evaluation set. The reported advantage is thus partly a product of the selection procedure, not an independent test of the method. Appendix C.1 similarly selects s=6 on the same dataset, making the reported configuration doubly fitted to the test set.

full rationale

LAI is a constructive pseudo-inverse: q is sampled so that argmax(q)=r, so perfect reconstruction holds by design and the paper states this openly; that is not a hidden circularity. The editing step q_t = p_t + (1−λ)g + λ n_t is a heuristic interpolation with no claimed derivation, so there is no equation-level equivalence. The only circularity-like issue is in the evaluation: τ=18 (and s=6) are selected by ablations on the PIE-Bench test benchmark (Appendix C.2, Table 4), and Table 1 then reports the exact τ=18 row as VARIN's headline result. Thus the reported advantage is partly in-sample selection rather than an independent prediction. Additionally, the Regeneration baseline in Table 1 is the s=8 configuration (matching Table 3), not the stated default s=6; at s=6 Regeneration has higher CLIP Edited (23.11 vs 21.49), undermining the 'adheres to target prompt more effectively' claim. This is a comparison-fairness/correctness issue, not circularity. The self-citations to DICE [17] are not load-bearing: LAI is a different pseudo-inverse and DICE is used as a baseline, not as a premise. Overall, the central construction is independent, but the headline quantitative claim is partially in-sample, so score 4.

Assumptions & free parameters 3 free parameters · 6 assumptions · 0 invented entities

The paper introduces no new physical or architectural entities. The central method rests on standard Gumbel-max machinery plus an empirically motivated interpolation scheme. The main free parameters are tuned on the evaluation dataset, which is the main circularity-in-evaluation risk.

free parameters (3)
  • tau = 18
    Information-preserving truncation margin in LAI. Chosen by ablation on the PIE-Bench evaluation set (Table 4); directly trades off background preservation against editability.
  • start scale s = 6
    First scale at which editing begins. Chosen by qualitative observation (Figure 2) and Regeneration ablation (Table 3, s=6 to 8); affects structure vs prompt alignment. Note the Regeneration baseline in Table 1 appears to use s=8 instead.
  • lambda scheduler = linear from 1 at scale s to 0 at scale 14
    Controls the blend of inverse noise vs fresh Gumbel noise. No ablation shown; set by hand (Section 5.2).
assumptions (6)
  • standard math Gumbel-max trick equivalence
    Sampling from a categorical distribution can be done by adding i.i.d. Gumbel noise to logits and taking the argmax; used throughout Section 3.1 and 4.1.
  • standard math Truncated Gumbel pseudo-inverse guarantees argmax
    Sampling non-label entries from Gumbel truncated above qmax - tau ensures the argmax remains at the label token; derived in Algorithm 1 and Algorithm 5.
  • domain assumption Next-scale AR factorization allows per-scale inversion
    Each scale token map depends only on previous scales, so inversion can be computed independently per scale (Equation 1, Algorithm 2). This is standard for AR models.
  • domain assumption Mid-scale editability
    Editing at scales 6 to 14 is assumed sufficient to realize prompt edits while coarse scales 1..5 preserve structure. Motivated by a single qualitative example (Figure 2).
  • ad hoc to paper Noise interpolation controls edit semantics
    The edit mechanism q_t = p_t + (1-lambda)*g + lambda*n_t is assumed to interpolate between source preservation and target-prompt generation. No derivation is provided (Section 4.3).
  • domain assumption Base model quality
    The method inherits the generation and VAE reconstruction quality of the base model (HART or Switti); the paper explicitly notes HART's weaker reconstruction vs Paella (Tables 2 and the discussion of Table 1).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Discrete Noise Inversion for Next-scale Autoregressive Text-based Image Editing." pith.science (2026). https://pith.science/paper/XM67MBKQ

@misc{pith2026250901984,
  author       = {Pith},
  title        = {Pith review of: Discrete Noise Inversion for Next-scale Autoregressive Text-based Image Editing},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XM67MBKQ}},
  note         = {Machine review of arXiv:2509.01984}
}
read the original abstract

Visual autoregressive models (VAR) have recently emerged as a promising class of generative models, achieving performance comparable to diffusion models in text-to-image generation tasks. While conditional generation has been widely explored, the ability to perform prompt-guided image editing without additional training is equally critical, as it supports numerous practical real-world applications. This paper investigates the text-to-image editing capabilities of VAR by introducing Visual AutoRegressive Inverse Noise (VARIN), the first noise inversion-based editing technique designed explicitly for VAR models. VARIN leverages a novel pseudo-inverse function for argmax sampling, named Location-aware Argmax Inversion (LAI), to generate inverse Gumbel noises. These inverse noises enable precise reconstruction of the source image and facilitate targeted, controllable edits aligned with textual prompts. Extensive experiments demonstrate that VARIN effectively modifies source images according to specified prompts while significantly preserving the original background and structural details, thus validating its efficacy as a practical editing approach.

Figures

Figures reproduced from arXiv: 2509.01984 by the authors.

Figure 1
Figure 1. Qualitative performance of VARIN given diverse prompts. [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Visualizations of each scale of the generation process of HART [ [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 4
Figure 4. To search for a more fine-grained editing control, we approach the noise inversion technique [PITH_FULL_IMAGE:figures/full_fig_p004_4.png] view at source ↗
Figures from the paper (9 more)
Figure 3
Figure 3. Figure 3: VARIN pipeline: the prefix token csrc + r<k is fed to transformer to get log probability pk. We then use pseudo inverse-argmax to find the inverse noise nk from ground truth label rk and logit pk. These noise set n1, n2, . . . , nK is later used for editing control. 4.…
Figure 4
Figure 4. Figure 4: Qualitative result of editing results between VARIN and baseline Regeneration. We should [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 6
Figure 6. Figure 6: Ablation of Regeneration on beginning scale for editing [PITH_FULL_IMAGE:figures/full_fig_p016_6.png]
Figure 7
Figure 7. Figure 7: Qualitative result for ablation of τ for VARIN. editing task. It is worth noting that for only target VARIN, the effective τ value is lower, typically between 10 and 14. In [PITH_FULL_IMAGE:figures/full_fig_p017_7.png]
Figure 8
Figure 8. Figure 8: The first column is source image. The second column is target VARIN Algorithm 6, and [PITH_FULL_IMAGE:figures/full_fig_p018_8.png]
Figure 9
Figure 9. Figure 9: Editing results on complex scene involves two objects. [PITH_FULL_IMAGE:figures/full_fig_p019_9.png]
Figure 10
Figure 10. Figure 10: Extending VARIN to different architectures and base models. [PITH_FULL_IMAGE:figures/full_fig_p019_10.png]
Figure 11
Figure 11. Figure 11: Comparing VARIN with different methods. Source Image Edited Image A goat and/kissing a cat A german shepherd dog standing/jumping [PITH_FULL_IMAGE:figures/full_fig_p020_11.png]
Figure 12
Figure 12. Figure 12: Failure cases: large movement and complex interaction between object [PITH_FULL_IMAGE:figures/full_fig_p020_12.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

66 extracted references · 34 canonical work pages

  1. [1]

    Achiam, S

    J. Achiam, S. Adler, S. Agarwal, L. Ahmad, I. Akkaya, F. L. Aleman, D. Almeida, J. Altenschmidt, S. Altman, S. Anadkat, et al. Gpt-4 technical report. arXiv preprint arXiv:2303.08774, 2023

    arXiv 2023

  2. [2]

    J. Bai, S. Bai, Y . Chu, Z. Cui, K. Dang, X. Deng, Y . Fan, W. Ge, Y . Han, F. Huang, et al. Qwen technical report. arXiv preprint arXiv:2309.16609, 2023

    arXiv 2023

  3. [3]

    Brack, F

    M. Brack, F. Friedrich, K. Kornmeier, L. Tsaban, P. Schramowski, K. Kersting, and A. Passos. Ledits++: Limitless image editing using text-to-image models. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , 2024

    work page 2024

  4. [4]

    Brooks, A

    T. Brooks, A. Holynski, and A. A. Efros. Instructpix2pix: Learning to follow image editing instructions. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 18392–18402, 2023

    2023

  5. [5]

    M. Cao, X. Wang, Z. Qi, Y . Shan, X. Qie, and Y . Zheng. Masactrl: Tuning-free mutual self- attention control for consistent image synthesis and editing. In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , pages 22560–22570, October 2023

    work page 2023

  6. [6]

    Chang, H

    H. Chang, H. Zhang, L. Jiang, C. Liu, and W. T. Freeman. Maskgit: Masked generative image transformer. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 11315–11325, 2022

    2022

  7. [7]

    R. T. Chen, Y . Rubanova, J. Bettencourt, and D. K. Duvenaud. Neural ordinary differential equations. Advances in neural information processing systems , 31, 2018

    2018

  8. [8]

    Chowdhery, S

    A. Chowdhery, S. Narang, J. Devlin, M. Bosma, G. Mishra, A. Roberts, P. Barham, H. W. Chung, C. Sutton, S. Gehrmann, et al. Palm: Scaling language modeling with pathways.Journal of Machine Learning Research, 24(240):1–113, 2023

    work page 2023

Show all 66 references
  1. [9]

    Dhariwal and A

    P. Dhariwal and A. Nichol. Diffusion models beat gans on image synthesis. Advances in neural information processing systems, 34:8780–8794, 2021

    2021

  2. [10]

    W. Dong, S. Xue, X. Duan, and S. Han. Prompt tuning inversion for text-driven image editing using diffusion models. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 7430–7440, 2023

    2023

  3. [11]

    Esser, R

    P. Esser, R. Rombach, A. Blattmann, and B. Ommer. Imagebart: Bidirectional context with multinomial diffusion for autoregressive image synthesis. Advances in neural information processing systems, 34:3518–3532, 2021

    2021

  4. [12]

    Esser, R

    P. Esser, R. Rombach, and B. Ommer. Taming transformers for high-resolution image synthesis. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages 12873–12883, 2021

    2021

  5. [13]

    T.-J. Fu, W. Hu, X. Du, W. Y . Wang, Y . Yang, and Z. Gan. Guiding instruction-based image editing via multimodal large language models. arXiv preprint arXiv:2309.17102, 2023

    2023 arXiv

  6. [14]

    Goodfellow, J

    I. Goodfellow, J. Pouget-Abadie, M. Mirza, B. Xu, D. Warde-Farley, S. Ozair, A. Courville, and Y . Bengio. Generative adversarial nets.Advances in neural information processing systems , 27, 2014

    2014

  7. [15]

    L. Han, Y . Li, H. Zhang, P. Milanfar, D. Metaxas, and F. Yang. Svdiff: Compact parameter space for diffusion fine-tuning. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 7323–7334, 2023

    2023

  8. [16]

    W. He, S. Fu, M. Liu, X. Wang, W. Xiao, F. Shu, Y . Wang, L. Zhang, Z. Yu, H. Li, et al. Mars: Mixture of auto-regressive models for fine-grained text-to-image synthesis. arXiv preprint arXiv:2407.07614, 2024

    2024 arXiv

  9. [17]

    X. He, L. Han, Q. Dao, S. Wen, M. Bai, D. Liu, H. Zhang, M. R. Min, F. Juefei-Xu, C. Tan, et al. Dice: Discrete inversion enabling controllable editing for multinomial diffusion and masked generative models. arXiv preprint arXiv:2410.08207, 2024. 11

    2024

  10. [18]

    Hertz, R

    A. Hertz, R. Mokady, J. Tenenbaum, K. Aberman, Y . Pritch, and D. Cohen-Or. Prompt-to- prompt image editing with cross attention control. arXiv preprint arXiv:2208.01626, 2022

    2022 arXiv

  11. [19]

    J. Ho, A. Jain, and P. Abbeel. Denoising diffusion probabilistic models. Advances in neural information processing systems, 33:6840–6851, 2020

    2020

  12. [20]

    Hoogeboom, D

    E. Hoogeboom, D. Nielsen, P. Jaini, P. Forré, and M. Welling. Argmax flows and multinomial diffusion: Learning categorical distributions. Advances in Neural Information Processing Systems, 34:12454–12465, 2021

    2021

  13. [21]

    Huang, J

    Y . Huang, J. Huang, Y . Liu, M. Yan, J. Lv, J. Liu, W. Xiong, H. Zhang, S. Chen, and L. Cao. Diffusion model-based image editing: A survey. arXiv preprint arXiv:2402.17525, 2024

    2024 arXiv

  14. [22]

    Huberman-Spiegelglas, V

    I. Huberman-Spiegelglas, V . Kulikov, and T. Michaeli. An edit friendly ddpm noise space: Inversion and manipulations. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 12469–12478, 2024

    2024

  15. [23]

    X. Ju, A. Zeng, Y . Bian, S. Liu, and Q. Xu. Direct inversion: Boosting diffusion-based editing with 3 lines of code. arXiv preprint arXiv:2310.01506, 2023

    2023 arXiv

  16. [24]

    D. P. Kingma, T. Salimans, R. Jozefowicz, X. Chen, I. Sutskever, and M. Welling. Improved vari- ational inference with inverse autoregressive flow. Advances in Neural Information Processing Systems, 29, 2016

    2016

  17. [25]

    W. Kool, H. Van Hoof, and M. Welling. Stochastic beams and where to find them: The gumbel-top-k trick for sampling sequences without replacement. In International Conference on Machine Learning, pages 3499–3508. PMLR, 2019

    2019

  18. [26]

    Kumari, B

    N. Kumari, B. Zhang, R. Zhang, E. Shechtman, and J.-Y . Zhu. Multi-concept customization of text-to-image diffusion. In CVPR, 2023

    2023

  19. [27]

    D. Lee, C. Kim, S. Kim, M. Cho, and W.-S. Han. Autoregressive image generation using residual quantization. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 11523–11532, 2022

    2022

  20. [28]

    T. Li, Y . Tian, H. Li, M. Deng, and K. He. Autoregressive image generation without vector quantization. arXiv preprint arXiv:2406.11838, 2024

    2024 arXiv

  21. [29]

    Lipman, R

    Y . Lipman, R. T. Chen, H. Ben-Hamu, M. Nickel, and M. Le. Flow matching for generative modeling. arXiv preprint arXiv:2210.02747, 2022

    2022 arXiv

  22. [30]

    X. Liu, C. Gong, and Q. Liu. Flow straight and fast: Learning to generate and transfer data with rectified flow. arXiv preprint arXiv:2209.03003, 2022

    2022 arXiv

  23. [31]

    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. arXiv preprint arXiv:2108.01073, 2021

    2021 arXiv

  24. [32]

    Miyake, A

    D. Miyake, A. Iohara, Y . Saito, and T. Tanaka. Negative-prompt inversion: Fast image inversion for editing with text-guided diffusion models. arXiv preprint arXiv:2305.16807, 2023

    2023 arXiv

  25. [33]

    Mokady, A

    R. Mokady, A. Hertz, K. Aberman, Y . Pritch, and D. Cohen-Or. Null-text inversion for editing real images using guided diffusion models. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 6038–6047, 2023

    2023

  26. [34]

    J. Mu, N. Vasconcelos, and X. Wang. Editar: Unified conditional generation with autoregressive models. arXiv preprint arXiv:2501.04699, 2025

    2025 arXiv

  27. [35]

    Nguyen, D.-A

    T.-T. Nguyen, D.-A. Nguyen, A. Tran, and C. Pham. Flexedit: Flexible and controllable diffusion-based object-centric image editing. arXiv preprint arXiv:2403.18605, 2024

    2024 arXiv

  28. [36]

    Poole, A

    B. Poole, A. Jain, J. T. Barron, and B. Mildenhall. Dreamfusion: Text-to-3d using 2d diffusion. arXiv, 2022. 12

    2022

  29. [37]

    Radford, J

    A. Radford, J. W. Kim, C. Hallacy, A. Ramesh, G. Goh, S. Agarwal, G. Sastry, A. Askell, P. Mishkin, J. Clark, et al. Learning transferable visual models from natural language supervision. In International conference on machine learning , pages 8748–8763. PMLR, 2021

    2021

  30. [38]

    Ramesh, M

    A. Ramesh, M. Pavlov, G. Goh, S. Gray, C. V oss, A. Radford, M. Chen, and I. Sutskever. Zero-shot text-to-image generation. In International conference on machine learning , pages 8821–8831. Pmlr, 2021

    2021

  31. [39]

    Rampas, P

    D. Rampas, P. Pernias, and M. Aubreville. A novel sampling scheme for text-and image- conditional image synthesis in quantized latent spaces. arXiv preprint arXiv:2211.07292 , 2022

    2022 arXiv

  32. [40]

    Razavi, A

    A. Razavi, A. Van den Oord, and O. Vinyals. Generating diverse high-fidelity images with vq-vae-2. Advances in neural information processing systems , 32, 2019

    2019

  33. [41]

    Rombach, A

    R. Rombach, A. Blattmann, D. Lorenz, P. Esser, and B. 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

    2022

  34. [42]

    N. Ruiz, Y . Li, V . Jampani, Y . Pritch, M. Rubinstein, and K. Aberman. Dreambooth: Fine tuning text-to-image diffusion models for subject-driven generation. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages 22500–22510, 2023

    2023

  35. [43]

    Sheynin, A

    S. Sheynin, A. Polyak, U. Singer, Y . Kirstain, A. Zohar, O. Ashual, D. Parikh, and Y . Taigman. Emu edit: Precise image editing via recognition and generation tasks. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 8871–8879, 2024

    2024

  36. [44]

    Y . Shi, C. Xue, J. H. Liew, J. Pan, H. Yan, W. Zhang, V . Y . Tan, and S. Bai. Dragdiffusion: Harnessing diffusion models for interactive point-based image editing. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 8839–8849, 2024

    2024

  37. [45]

    J. Song, C. Meng, and S. Ermon. Denoising diffusion implicit models. In International Conference on Learning Representations, 2021

    2021

  38. [46]

    Y . Song, J. Sohl-Dickstein, D. P. Kingma, A. Kumar, S. Ermon, and B. Poole. Score-based generative modeling through stochastic differential equations. arXiv preprint arXiv:2011.13456, 2020

    2011 arXiv

  39. [47]

    P. Sun, Y . Jiang, S. Chen, S. Zhang, B. Peng, P. Luo, and Z. Yuan. Autoregressive model beats diffusion: Llama for scalable image generation. arXiv preprint arXiv:2406.06525, 2024

    2024 arXiv

  40. [48]

    H. Tang, Y . Wu, S. Yang, E. Xie, J. Chen, J. Chen, Z. Zhang, H. Cai, Y . Lu, and S. Han. Hart: Effi- cient visual generation with hybrid autoregressive transformer.arXiv preprint arXiv:2410.10812, 2024

    2024 arXiv

  41. [49]

    G. Team, R. Anil, S. Borgeaud, J.-B. Alayrac, J. Yu, R. Soricut, J. Schalkwyk, A. M. Dai, A. Hauth, K. Millican, et al. Gemini: a family of highly capable multimodal models. arXiv preprint arXiv:2312.11805, 2023

    2023 arXiv

  42. [50]

    K. Tian, Y . Jiang, Z. Yuan, B. Peng, and L. Wang. Visual autoregressive modeling: Scalable image generation via next-scale prediction. arXiv preprint arXiv:2404.02905, 2024

    2024 arXiv

  43. [51]

    Touvron, T

    H. Touvron, T. Lavril, G. Izacard, X. Martinet, M.-A. Lachaux, T. Lacroix, B. Rozière, N. Goyal, E. Hambro, F. Azhar, et al. Llama: Open and efficient foundation language models. arXiv preprint arXiv:2302.13971, 2023

    2023 arXiv

  44. [52]

    Tumanyan, M

    N. Tumanyan, M. Geyer, S. Bagon, and T. Dekel. Plug-and-play diffusion features for text- driven image-to-image translation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 1921–1930, 2023

    1921

  45. [53]

    Van Den Oord, O

    A. Van Den Oord, O. Vinyals, et al. Neural discrete representation learning. Advances in neural information processing systems, 30, 2017. 13

    2017

  46. [54]

    Van Le, H

    T. Van Le, H. Phung, T. H. Nguyen, Q. Dao, N. N. Tran, and A. Tran. Anti-dreambooth: Protecting users from personalized text-to-image synthesis. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 2116–2127, 2023

    2023

  47. [55]

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

    2017

  48. [56]

    V oronov, D

    A. V oronov, D. Kuznedelev, M. Khoroshikh, V . Khrulkov, and D. Baranchuk. Switti: Designing scale-wise transformers for text-to-image synthesis. 2024

    2024

  49. [57]

    Z. Wang, A. C. Bovik, H. R. Sheikh, and E. P. Simoncelli. Image quality assessment: from error visibility to structural similarity. IEEE transactions on image processing , 13(4):600–612, 2004

    2004

  50. [58]

    Z. Wang, C. Lu, Y . Wang, F. Bao, C. Li, H. Su, and J. Zhu. Prolificdreamer: High-fidelity and diverse text-to-3d generation with variational score distillation.arXiv preprint arXiv:2305.16213, 2023

    2023 arXiv

  51. [59]

    Workshop, T

    B. Workshop, T. L. Scao, A. Fan, C. Akiki, E. Pavlick, S. Ili´c, D. Hesslow, R. Castagné, A. S. Luccioni, F. Yvon, et al. Bloom: A 176b-parameter open-access multilingual language model. arXiv preprint arXiv:2211.05100, 2022

    2022 arXiv

  52. [60]

    C. H. Wu and F. De la Torre. Unifying diffusion models’ latent space, with applications to cyclediffusion and guidance. arXiv preprint arXiv:2210.05559, 2022

    2022 arXiv

  53. [61]

    C. H. Wu and F. D. la Torre. A latent space of stochastic diffusion models for zero-shot image editing and guidance. In ICCV, 2023

    2023

  54. [62]

    L. Yu, J. Lezama, N. B. Gundavarapu, L. Versari, K. Sohn, D. Minnen, Y . Cheng, V . Birodkar, A. Gupta, X. Gu, et al. Language model beats diffusion–tokenizer is key to visual generation. arXiv preprint arXiv:2310.05737, 2023

    2023 arXiv

  55. [63]

    Zhang, A

    L. Zhang, A. Rao, and M. Agrawala. Adding conditional control to text-to-image diffusion models. In Proceedings of the IEEE/CVF International Conference on Computer Vision , pages 3836–3847, 2023

    2023

  56. [64]

    Zhang, P

    R. Zhang, P. Isola, A. A. Efros, E. Shechtman, and O. Wang. The unreasonable effectiveness of deep features as a perceptual metric. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 586–595, 2018

    2018

  57. [65]

    Zhang, X

    S. Zhang, X. Yang, Y . Feng, C. Qin, C.-C. Chen, N. Yu, Z. Chen, H. Wang, S. Savarese, S. Ermon, et al. Hive: Harnessing human feedback for instructional visual editing. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 9026–9036, 2024

    2024

  58. [66]

    Zhang, L

    Z. Zhang, L. Han, A. Ghosh, D. N. Metaxas, and J. Ren. Sine: Single image editing with text-to-image diffusion models. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 6027–6037, 2023. 14 A Appendix In this supplementary material, we ...

    2023

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.