Pith. sign in

REVIEW 4 major objections 6 minor 1 cited by

Normalized Attention Guidance: Universal Negative Guidance for Diffusion Models

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 Attention-space trick revives negative guidance in few-step diffusion

desk verdict A genuinely useful stabilization of attention-space negative guidance, but the paper's central claim that it provides semantic negative guidance is not yet isolated from generic quality enhancement. read the letter →

arxiv 2505.21179 v3 pith:3H4R47TT submitted 2025-05-27 cs.CV

Dar-Yen Chen , Hmrishav Bandyopadhyay , Kai Zou , Yi-Zhe Song This is my paper
classification cs.CV
keywords diffusionmodelsnegativepromptingclassifier-freeguidanceattentionfew-stepsamplingNormalizedDiTvideogeneration
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 claims that negative prompting—asking a diffusion model to suppress unwanted attributes such as "blurry" or "glasses"—can be made reliable in fast few-step samplers, where standard classifier-free guidance (CFG) breaks down because the positive and negative predictions diverge. It introduces Normalized Attention Guidance (NAG), a training-free plug-in that steers generation by extrapolating the cross-attention outputs away from the negative prompt, then applies L1-norm clipping and alpha-blending to keep those features near the valid data manifold. If true, this would restore a core controllability tool for high-efficiency diffusion models and extend negative prompting to new domains without retraining.

What carries the argument

The central object is the NAG feature transform applied inside cross-attention layers: $\tilde{Z} = Z^{+} + \phi (Z^{+} - Z^{-})$, followed by token-wise L1-norm ratio clipping, $\hat{Z}[i] = \frac{\min(R[i], \tau)}{R[i]} \tilde{Z}[i]$ with $R[i] = \frac{\| \tilde{Z}[i] \|_1}{\| Z^{+}[i] \|_1}$, and final refinement $Z_{\mathrm{NAG}} = \alpha \hat{Z} + (1-\alpha) Z^{+}$. This transform replaces output-space CFG extrapolation with a bounded, direction-preserving move in attention space; the two stabilization operations work as a magnitude guardrail and a manifold anchor, respectively, and are what let the guidance remain stable at high scale.

What would settle it

Reproduce Table 1 on 4-step Flux-Schnell with the universal negative prompt 'Low resolution, blurry'. The claim would be falsified if (a) the reported CLIP, FID, and ImageReward improvements do not reproduce, (b) ablations without normalization or without blending match or beat the full method, or (c) at a fixed scale the extrapolated attention features $\hat{Z}$ land measurably far outside the distribution of unmodified $Z^{+}$ in a way that predicts the observed artifacts.

Watch

Extended reading notes

Core claim

The paper's central claim is that effective negative guidance can be recovered in few-step diffusion models by moving extrapolation out of the predicted-noise/output space and into the attention feature space, with two stabilizing operations that prior attention-manipulation methods like NASA lack. Specifically, NAG computes attention outputs $Z^{+}$ and $Z^{-}$ from the positive and negative prompts, forms an extrapolated feature $\tilde{Z} = Z^{+} + \phi (Z^{+} - Z^{-})$, then rescales each token's L1 norm to a bounded threshold $\tau$ and blends the result with $Z^{+}$ by a factor $\alpha$. The authors argue the L1 normalization preserves low-magnitude activations that carry subtle semantics while constraining extremes, and that the blending pulls features back toward the stable positive manifold; together these prevent the out-of-manifold drift and feature collapse they observe in prior methods. They claim this makes NAG a universal plug-in: it works where CFG collapses (few-step UNet and DiT models such as Flux-Schnell, SDXL-Lightning, and DMD2-SDXL), improves multi-step models when added on top of CFG and PAG, and transfers to video diffusion (Wan2.1) for both content suppression and motion enhancement.

Load-bearing premise

The whole method rests on the heuristic that L1-norm clipping plus alpha-blending keeps extrapolated attention features close enough to the true feature manifold, and that this holds across architectures and modalities without per-model tuning; the paper does not derive this from theory, only from the demonstrated experiments.

Editorial extensions

If this is right

  • Few-step models that ship without CFG support (Flux-Schnell, SANA-Sprint, SD3.5-Large-Turbo, NitroSD-Realism, DMD2-SDXL, SDXL-Lightning) gain a working negative-prompting interface without retraining.
  • When combined with CFG and PAG in 25-step models, NAG improves CLIP score and ImageReward while leaving FID and PFID roughly flat or slightly better, giving a way to add negative control on top of existing guidance.
  • Applying NAG only for the first fraction of denoising steps (e.g., $\theta = 0.25$) keeps most of the measured benefit while cutting the added latency, since the guidance effect is concentrated in early timesteps.
  • NAG extends to video generation on Wan2.1, suppressing undesired content such as human faces or static motion and improving temporal dynamics.
  • The method requires no retraining and adds less latency than CFG on most model families (e.g., +13% on SANA and +12% on Wan2.1 versus +90% to +100% for CFG), making it a practical drop-in at inference time.

Reading between the lines

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

  • The consistent behavior under a fixed L1 clip and blend across very different architectures hints that attention features in these models share a common scale structure; one could test this directly by measuring the distribution of $\|Z\|_1$ across layers and models and checking whether the default $\tau = 2.5$ tracks a consistent quantile.
  • Because NAG operates on frozen attention features rather than on training, the same mechanism could be aimed at other steering objectives—concept erasure, style transfer, or prompt-based debiasing—provided a negative description can be written; the paper demonstrates the prompting use, not the full space of guidance tasks.
  • The early-stopping result suggests the semantic direction of a negative prompt is mostly determined in the first denoising steps; if bias suppression is applied only there, downstream steps remain untouched, which might make NAG-based debiasing less likely to degrade fine texture.
  • Nothing in the method requires the negative condition to be textual; using an image as the negative condition could test whether NAG's attention-space extrapolation works for reference-based editing.
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 Normalized Attention Guidance (NAG), a training-free inference-time method for negative prompting in diffusion models. NAG computes attention outputs for positive and negative prompts, extrapolates in attention feature space as eZ = Z+ + φ(Z+ - Z-), applies L1-ratio clipping with threshold τ, and blends with the positive output using α. The authors claim that NAG restores effective negative guidance where CFG collapses, particularly in few-step models, and that it generalizes across UNet/DiT architectures, image/video modalities, and few-/multi-step settings. Experiments on seven few-step models, several multi-step CFG models, and one video model report CLIP/FID/PFID/ImageReward gains, supplemented by ablations, latency measurements, and a user study.

Significance. If the causal attribution to negative-attribute suppression is established, NAG would be a practical and broadly applicable plug-in for diffusion models. The method is simple and clearly specified (Algorithm 1, Table 5), and the paper provides unusually broad model coverage, including recent DiT models and a 14B video model, along with latency measurements and a component ablation. The authors are also candid about failure cases in Section 6. However, the current evaluation does not isolate negative-attribute suppression from generic quality enhancement, so the central claim is not yet verified. The paper also lacks comparisons with standard negative-prompt baselines and statistical uncertainty estimates, which are necessary to support the 'universal' and 'consistent improvements' claims.

major comments (4)
  1. [§5.1, Table 1, §5.5] The quantitative evaluation relies on the single negative prompt 'Low resolution, blurry' and on aggregate quality metrics (CLIP, FID, PFID, ImageReward) that would also improve from generic sharpening or contrast enhancement. The paper does not report any direct measure of the presence of the suppressed attribute (e.g., classifier accuracy for 'glasses', 'tiger', or 'male', or CLIP similarity to the negative text), and the user study in Section 5.5 instructs participants to judge text alignment while considering both positive and negative prompts, so it cannot separate suppression from aesthetic improvement. Consequently, the central claim that NAG implements directional negative guidance rather than acting as a generic feature regularizer is not yet supported. I recommend adding attribute-presence metrics on the generated samples for several negative concepts, or a controlled experiment in which the negative prompt targets a semantic attribute while the positive prompt does not mention it.
  2. [§5.1–5.3] There is no baseline against standard text-level negative prompting. For multi-step models, CFG with a negative prompt is the default practice; for few-step models, straightforward baselines such as prompt inversion or simply appending the negative prompt to the positive prompt could be used. Without such comparisons, the gains in Tables 1 and 2 could stem from the additional negative conditioning per se rather than from the attention-space mechanism of NAG. Please add these baselines to the evaluation.
  3. [Table 5, Table 2, Table 6] The hyperparameters (ϕ, τ, α) in Table 5 are chosen per model family, and the paper does not state whether they were tuned on the same COCO-5K evaluation set used in Tables 1, 2, and 6. Moreover, no error bars or repeated runs are reported; several FID and PFID differences are within typical noise (e.g., SD3.5-Large CFG FID +0.35 in Table 2, SD1.5 FID +0.14 and PFID +1.29 in Table 6). This weakens the 'consistent improvements' and 'universal' claims. Please report the tuning protocol and provide confidence intervals or at least per-prompt variance.
  4. [§5.2, Appendix E] The comparison with NASA is limited to UNet models, and the user study against NASA (Table 8) is conducted on only one model (DMD2-SDXL). Since the paper motivates NAG partly by NASA's instability on DiT architectures, a quantitative DiT comparison (even if it shows NASA collapse) would strengthen the claim. Also, the attribute-specific suppression of NAG versus NASA is not compared directly, so it remains unclear whether NAG's advantage is in guidance direction or in overall stability.
minor comments (6)
  1. [Algorithm 1] The computed z_nag is never assigned to hidden_states; the function returns the unmodified hidden_states. Replace 'return hidden_states' with 'return z_nag' or add an assignment before the return.
  2. [Equations (8)–(9)] The symbol R is used both for the per-token ratio R[i] and for the vector in Eq. (9); please use a different symbol for the vector (e.g., lowercase r) and make the element-wise division explicit.
  3. [References and Table 2] Table 2 and the main text cite PAG as [12], but [12] is 'Rectified diffusion guidance' whereas Perturbed Attention Guidance is [59]; please correct the citation.
  4. [Conclusion] There is a typo: 'the the limitations' should be 'the limitations,' and 'CFG ... fails' should agree in number with 'CFG'.
  5. [Figure 4] The terms 'Guidance Boundary' and 'Refinement Manifold' are introduced only in the caption; please define them in the text or make the figure self-contained.
  6. [Section 5.4] The video evaluation is qualitative only; the contribution statement in Section 1 about improving 'motion characteristics' would benefit from a quantitative metric on the generated videos.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: NAG is an empirical, measured method; no claimed derivation reduces to its inputs.

full rationale

This paper proposes an inference-time heuristic (attention-space extrapolation with L1 normalization and α-blending) and evaluates it with standard generative metrics. There is no derivation chain in which a predicted quantity is defined in terms of a fitted quantity or in which an output is equivalent to an input by construction. The method's components—extrapolation (Eq. 7), normalization (Eqs. 8–9), and refinement (Eq. 10)—are stated as operations, not derived from the metrics they are judged against. The claims of restored negative guidance are supported by measured CLIP/FID/PFID/ImageReward changes and user studies; these are empirical observations rather than consequences of the method's definitions. The few self-citations (e.g., Nitrofusion as an evaluation backbone) are not load-bearing for the central claim. The choice of the universal negative prompt 'Low resolution, blurry' and per-model-family hyperparameters may raise evaluation-validity questions (e.g., in-sample tuning or quality-enhancement confounds), but those are not circular reasoning: the reported numbers are not forced by the equations. No circular step can be quoted or exhibited, so the appropriate finding is no significant circularity.

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

The method introduces two per-family hyperparameters (phi, alpha) and a fixed threshold (tau). It also relies on several domain assumptions about attention feature geometry that are not proven.

free parameters (3)
  • Guidance scale phi = Per model family: Flux 4, SD3.5-Large 4, SANA 4, PixArt-Sigma 4, Wan2.1 4, SDXL 2, Playground 2, SD1.5 2
    Tuned per model family in Table 5; controls extrapolation strength.
  • Refinement factor alpha = Per model family: Flux 0.25, SD3.5-Large 0.125, SANA 0.375, PixArt-Sigma 0.375, Wan2.1 0.25, SDXL 0.5, Playground 0.5…
    Tuned per model family in Table 5; controls blending with positive baseline.
  • Norm threshold tau = 2.5 for all model families
    Fixed by hand; ablation in Appendix F shows sensitivity, but the paper uses 2.5 universally.
assumptions (4)
  • domain assumption Cross-attention features encode semantic information in the direction of the prompt.
    Required for the claim that moving attention features away from Z- suppresses undesired semantics.
  • domain assumption L1 norm preserves low-magnitude activations that encode subtle semantics better than L2.
    Used to justify the choice of L1 normalization in Section 4.1.
  • domain assumption Blending with the positive baseline pulls features back to the data manifold.
    The refinement step (Eq. 10) assumes this regularizes out-of-manifold drift.
  • domain assumption The pre-trained diffusion model is well-calibrated enough to be guided by attention feature perturbations.
    The entire method relies on the idea that modifying cross-attention outputs does not destroy generation quality.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Normalized Attention Guidance: Universal Negative Guidance for Diffusion Models." pith.science (2026). https://pith.science/paper/3H4R47TT

@misc{pith2026250521179,
  author       = {Pith},
  title        = {Pith review of: Normalized Attention Guidance: Universal Negative Guidance for Diffusion Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3H4R47TT}},
  note         = {Machine review of arXiv:2505.21179}
}
read the original abstract

Negative guidance -- explicitly suppressing unwanted attributes -- remains a fundamental challenge in diffusion models, particularly in few-step sampling regimes. While Classifier-Free Guidance (CFG) works well in standard settings, it fails under aggressive sampling step compression due to divergent predictions between positive and negative branches. We present Normalized Attention Guidance (NAG), an efficient, training-free mechanism that applies extrapolation in attention space with L1-based normalization and refinement. NAG restores effective negative guidance where CFG collapses while maintaining fidelity. Unlike existing approaches, NAG generalizes across architectures (UNet, DiT), sampling regimes (few-step, multi-step), and modalities (image, video), functioning as a \textit{universal} plug-in with minimal computational overhead. Through extensive experimentation, we demonstrate consistent improvements in text alignment (CLIP Score), fidelity (FID, PFID), and human-perceived quality (ImageReward). Our ablation studies validate each design component, while user studies confirm significant preference for NAG-guided outputs. As a model-agnostic inference-time approach requiring no retraining, NAG provides effortless negative guidance for all modern diffusion frameworks -- pseudocode in the Appendix!

Figures

Figures reproduced from arXiv: 2505.21179 by the authors.

Figure 1
Figure 1. Negative prompting on 4-step Flux-Schnell [1]. CFG fails in few-step models. NAG restores effective negative prompting, enabling direct suppression of visual, semantic, and stylistic attributes, such as “glasses,” “tiger,” “realistic,” or “blurry.” This enhances controllability and expands creative freedom across composition, style, and quality—including prompt-based debiasing. Preprint. Under review. arXiv:2505.211… view at source ↗
Figure 2
Figure 2. Comparison of NAG against NASA. CFG [9] applies an extrapolation on noise pre￾dictions, which can be rewritten in terms of x0 by substituting Eq. (2) into Eq. (3): x CFG 0 = x + 0 + ϕ · (x + 0 − x − 0 ), (6) where x + 0 and x − 0 are reconstructed from posi￾tive and negative conditions, respectively. CFG inherently assumes a multi-step denoising pro￾cess to keep these branches aligned, where sam￾ples are gradually r… view at source ↗
Figure 3
Figure 3. Comparison of CFG and NAG in single-step sampling. Left: Classifier-Free Guidance (CFG) [9] generates x + 0 and x − 0 from positive and negative prompts, then applies output-space extrapolation. In few-step models, x + 0 and x − 0 differ significantly due to coarse denoising, leading to severe artifacts rather than controlled guidance. Right: Normalized Attention Guidance (NAG) operates in attention space by extrapo… view at source ↗
Figures from the paper (19 more)
Figure 4
Figure 4. Figure 4: Visualization of NAG. Feature Refinement. Though normalization constrains magni￾tude, it may still disrupt alignment with the original distribution. To mitigate this, we blend Zb with the positive baseline: Z NAG = α · Zb + (1 − α) · Z +. (10) This blending serves as a…
Figure 5
Figure 5. Figure 5: Qualitative results of NAG. NAG enhances controllability in models lacking CFG, improving semantic alignment and visual quality across architectures and sampling steps. 6 [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: NAG integration with CFG and PAG. To evaluate compatibility with existing guid￾ance, we integrate NAG into 25-step SD3.5- Large [15] and SDXL [61] models alongside CFG [9] and PAG [12] in [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: Qualitative video results for Wan2.1-T2V-14B. 7 [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]
Figure 8
Figure 8. Figure 8: Left: Ablation study of NAG. Right: Visual results from ablation study. Impact of guidance scale. We evaluate NAG’s performance under varying guidance scales in [PITH_FULL_IMAGE:figures/full_fig_p008_8.png]
Figure 9
Figure 9. Figure 9: Impact of NAG scale. =0 =8 =16 31.5 32.0 32.5 CLIP ( ) =0 =8 =16 23 24 25 FID ( ) =0 =8 =16 1.00 1.05 1.10 ImageReward ( ) 1 step 4 steps [PITH_FULL_IMAGE:figures/full_fig_p009_9.png]
Figure 10
Figure 10. Figure 10: Quantitative comparison of NAG scale [PITH_FULL_IMAGE:figures/full_fig_p009_10.png]
Figure 11
Figure 11. Figure 11: Examples of NAG failure cases. While NAG exhibits effective negative guid￾ance, some failure cases remain. As shown in [PITH_FULL_IMAGE:figures/full_fig_p009_11.png]
Figure 12
Figure 12. Figure 12: Additional qualitative results of NAG. 15 [PITH_FULL_IMAGE:figures/full_fig_p015_12.png]
Figure 13
Figure 13. Figure 13: Left: Impact of τ . Right: Impact of α. 16 [PITH_FULL_IMAGE:figures/full_fig_p016_13.png]
Figure 14
Figure 14. Figure 14: Visualization of the sampling process with NAG. 17 [PITH_FULL_IMAGE:figures/full_fig_p017_14.png]
Figure 15
Figure 15. Figure 15: Qualitative results of early stopping [PITH_FULL_IMAGE:figures/full_fig_p018_15.png]
Figure 16
Figure 16. Figure 16: User preference study interface. Images and videos are presented in a random order. 19 [PITH_FULL_IMAGE:figures/full_fig_p019_16.png]
Figure 17
Figure 17. Figure 17: Additional qualitative results of NAG for text-to-video generation. 20 [PITH_FULL_IMAGE:figures/full_fig_p020_17.png]
Figure 18
Figure 18. Figure 18: Qualitative results of NAG for image-to-video. 21 [PITH_FULL_IMAGE:figures/full_fig_p021_18.png]
Figure 19
Figure 19. Figure 19: Detailed qualitative analysis of Flux-Dev [1] and SD3.5-Large-Turbo [15, 52]. 22 [PITH_FULL_IMAGE:figures/full_fig_p022_19.png]
Figure 20
Figure 20. Figure 20: Detailed qualitative analysis of Flux-Schnell [1] and SANA-Sprint [53]. 23 [PITH_FULL_IMAGE:figures/full_fig_p023_20.png]
Figure 21
Figure 21. Figure 21: Uncurated qualitative results for Flux-Dev [1] and SD3.5-Large-Turbo [15, 52]. All prompts are detailed in Appendix J.1. NAG-guided sample utilizes the negative prompt “Low￾resolution, blurry.” 24 [PITH_FULL_IMAGE:figures/full_fig_p024_21.png]
Figure 22
Figure 22. Figure 22: Uncurated qualitative results for Flux-Schnell [1] and SANA-Sprint [53]. All prompts are detailed in Appendix J.1. NAG-guided sample utilizes the negative prompt “Low-resolution, blurry.” 25 [PITH_FULL_IMAGE:figures/full_fig_p025_22.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. MissingBench-Verified: Probing Vision-Language Models' Inability to Detect Missing Object Parts

    cs.CV 2026-07 conditional novelty 5.0 of 10

    Ten leading VLMs mostly fail to report removed essential object parts as missing, and simulated detector evidence, image tools, longer reasoning, and an easier fine-tune barely improve accuracy.

Reference graph

Works this paper leans on

87 extracted references · 44 canonical work pages · cited by 1 Pith paper

  1. [1]

    Black Forest Labs. Flux. https://github.com/black-forest-labs/flux , 2024

    2024

  2. [2]

    High-resolution image synthesis with latent diffusion models

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Björn Ommer. High-resolution image synthesis with latent diffusion models. In CVPR, 2022

    2022

  3. [3]

    Denoising diffusion probabilistic models

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising diffusion probabilistic models. In NeurIPS, 2020

    2020

  4. [4]

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

    Chitwan Saharia, William Chan, Saurabh Saxena, Lala Li, Jay Whang, Emily L Denton, Kamyar Ghasemipour, Raphael Gontijo Lopes, Burcu Karagol Ayan, Tim Salimans, et al. Photorealistic text-to- image diffusion models with deep language understanding. In NeurIPS, 2022

    2022

  5. [5]

    Kingma, Ben Poole, Mohammad Norouzi, David J

    Jonathan Ho, William Chan, Chitwan Saharia, Jay Whang, Ruiqi Gao, Alexey Gritsenko, Diederik P. Kingma, Ben Poole, Mohammad Norouzi, David J. Fleet, and Tim Salimans. Imagen video: High definition video generation with diffusion models. arXiv preprint arXiv:2210.02303, 2022

    arXiv

  6. [6]

    Wan: Open and advanced large-scale video generative models

    Team Wan, Ang Wang, Baole Ai, Bin Wen, Chaojie Mao, Chen-Wei Xie, Di Chen, Feiwu Yu, Haiming Zhao, Jianxiao Yang, Jianyuan Zeng, Jiayu Wang, Jingfeng Zhang, Jingren Zhou, Jinkai Wang, Jixuan Chen, Kai Zhu, Kang Zhao, Keyu Yan, Lianghua Huang, Mengyang Feng, Ningyi Zhang, Pandeng Li, Pingyu Wu, Ruihang Chu, Ruili Feng, Shiwei Zhang, Siyang Sun, Tao Fang, T...

    arXiv 2025

  7. [7]

    Sdxl-lightning: Progressive adversarial diffusion distillation

    Shanchuan Lin, Anran Wang, and Xiao Yang. Sdxl-lightning: Progressive adversarial diffusion distillation. arXiv preprint arXiv:2402.13929, 2024

    arXiv 2024

  8. [8]

    Nitrofusion: High-fidelity single- step diffusion through dynamic adversarial training

    Dar-Yen Chen, Hmrishav Bandyopadhyay, Kai Zou, and Yi-Zhe Song. Nitrofusion: High-fidelity single- step diffusion through dynamic adversarial training. arXiv preprint arXiv:2412.02030, 2024

    arXiv 2024

Show all 87 references
  1. [9]

    Classifier-free diffusion guidance

    Jonathan Ho and Tim Salimans. Classifier-free diffusion guidance. arXiv preprint arXiv:2207.12598 , 2022

    2022 arXiv

  2. [10]

    Eliminating oversaturation and artifacts of high guidance scales in diffusion models

    Seyedmorteza Sadat, Otmar Hilliges, and Romann M Weber. Eliminating oversaturation and artifacts of high guidance scales in diffusion models. In ICLR, 2024

    2024

  3. [11]

    Characteristic guidance: Non-linear correction for diffusion model at large guidance scale

    Candi Zheng and Yuan Lan. Characteristic guidance: Non-linear correction for diffusion model at large guidance scale. arXiv preprint arXiv:2312.07586, 2023

    2023 arXiv

  4. [12]

    Rectified diffusion guidance for conditional generation

    Mengfei Xia, Nan Xue, Yujun Shen, Ran Yi, Tieliang Gong, and Yong-Jin Liu. Rectified diffusion guidance for conditional generation. arXiv preprint arXiv:2410.18737, 2024

    2024

  5. [13]

    Snoopi: Supercharged one-step diffusion distillation with proper guidance

    Viet Nguyen, Anh Nguyen, Trung Dao, Khoi Nguyen, Cuong Pham, Toan Tran, and Anh Tran. Snoopi: Supercharged one-step diffusion distillation with proper guidance. arXiv preprint arXiv:2412.02687, 2024

    2024

  6. [14]

    Scalable diffusion models with transformers

    William Peebles and Saining Xie. Scalable diffusion models with transformers. In ICCV, 2023

    2023

  7. [15]

    Scaling rectified flow transformers for high-resolution image synthesis

    Patrick Esser, Sumith Kulal, Andreas Blattmann, Rahim Entezari, Jonas Müller, Harry Saini, Yam Levi, Dominik Lorenz, Axel Sauer, Frederic Boesel, et al. Scaling rectified flow transformers for high-resolution image synthesis. In ICML, 2024

    2024

  8. [16]

    Learning transferable visual models from natural language supervision

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning transferable visual models from natural language supervision. In ICML, 2021

    2021

  9. [17]

    Gans trained by a two time-scale update rule converge to a local nash equilibrium

    Martin Heusel, Hubert Ramsauer, Thomas Unterthiner, Bernhard Nessler, and Sepp Hochreiter. Gans trained by a two time-scale update rule converge to a local nash equilibrium. In NeurIPS, 2017

    2017

  10. [18]

    Imagereward: Learning and evaluating human preferences for text-to-image generation

    Jiazheng Xu, Xiao Liu, Yuchen Wu, Yuxuan Tong, Qinkai Li, Ming Ding, Jie Tang, and Yuxiao Dong. Imagereward: Learning and evaluating human preferences for text-to-image generation. In NeurIPS, 2024

    2024

  11. [19]

    Diffusion models beat gans on image synthesis

    Prafulla Dhariwal and Alexander Nichol. Diffusion models beat gans on image synthesis. In NeurIPS, 2021

    2021

  12. [20]

    Hierarchical text-conditional image generation with clip latents

    Aditya Ramesh, Prafulla Dhariwal, Alex Nichol, Casey Chu, and Mark Chen. Hierarchical text-conditional image generation with clip latents. arXiv preprint arXiv:2204.06125, 2022

    2022 arXiv

  13. [21]

    ediff-i: Text-to-image diffusion models with an ensemble of expert denoisers

    Yogesh Balaji, Seungjun Nah, Xun Huang, Arash Vahdat, Jiaming Song, Qinsheng Zhang, Karsten Kreis, Miika Aittala, Timo Aila, Samuli Laine, Bryan Catanzaro, Tero Karras, and Ming-Yu Liu. ediff-i: Text-to-image diffusion models with an ensemble of expert denoisers. arXiv preprin...

    2022 arXiv

  14. [22]

    Make-a-video: Text-to-video generation without text-video data

    Uriel Singer, Adam Polyak, Thomas Hayes, Xi Yin, Jie An, Songyang Zhang, Qiyuan Hu, Harry Yang, Oron Ashual, Oran Gafni, Devi Parikh, Sonal Gupta, and Yaniv Taigman. Make-a-video: Text-to-video generation without text-video data. arXiv preprint arXiv:2209.14792, 2022. 10

    2022 arXiv

  15. [23]

    Structure and content-guided video synthesis with diffusion models

    Patrick Esser, Johnathan Chiu, Parmida Atighehchian, Jonathan Granskog, and Anastasis Germanidis. Structure and content-guided video synthesis with diffusion models. In ICCV, 2023

    2023

  16. [24]

    Align your latents: High-resolution video synthesis with latent diffusion models

    Andreas Blattmann, Robin Rombach, Huan Ling, Tim Dockhorn, Seung Wook Kim, Sanja Fidler, and Karsten Kreis. Align your latents: High-resolution video synthesis with latent diffusion models. In CVPR, 2023

    2023

  17. [25]

    Photorealistic video generation with diffusion models

    Agrim Gupta, Lijun Yu, Kihyuk Sohn, Xiuye Gu, Meera Hahn, Li Fei-Fei, Irfan Essa, Lu Jiang, and José Lezama. Photorealistic video generation with diffusion models. In ECCV, 2024

    2024

  18. [26]

    Score-based generative modeling through stochastic differential equations

    Yang Song, Jascha Sohl-Dickstein, Diederik P Kingma, Abhishek Kumar, Stefano Ermon, and Ben Poole. Score-based generative modeling through stochastic differential equations. In ICLR, 2021

    2021

  19. [27]

    Elucidating the design space of diffusion-based generative models

    Tero Karras, Miika Aittala, Timo Aila, and Samuli Laine. Elucidating the design space of diffusion-based generative models. In NeurIPS, 2022

    2022

  20. [28]

    Building normalizing flows with stochastic interpolants

    Michael Samuel Albergo and Eric Vanden-Eijnden. Building normalizing flows with stochastic interpolants. In ICLR, 2023

    2023

  21. [29]

    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 ICLR, 2023

    2023

  22. [30]

    Albergo, Nicholas M

    Nanye Ma, Mark Goldstein, Michael S. Albergo, Nicholas M. Boffi, Eric Vanden-Eijnden, and Saining Xie. Sit: Exploring flow and diffusion-based generative models with scalable interpolant transformers. In ECCV, 2024

    2024

  23. [31]

    Improving and generalizing flow-based generative models with minibatch optimal transport

    Alexander Tong, Kilian FATRAS, Nikolay Malkin, Guillaume Huguet, Yanlei Zhang, Jarrid Rector-Brooks, Guy Wolf, and Yoshua Bengio. Improving and generalizing flow-based generative models with minibatch optimal transport. TMLR, 2024

    2024

  24. [32]

    Aram-Alexandre Pooladian, Heli Ben-Hamu, Carles Domingo-Enrich, Brandon Amos, Yaron Lipman, and Ricky T. Q. Chen. Multisample flow matching: Straightening flows with minibatch couplings. arXiv preprint arXiv:2304.14772, 2023

    2023 arXiv

  25. [33]

    Flow matching in latent space

    Quan Dao, Hao Phung, Binh Nguyen, and Anh Tran. Flow matching in latent space. arXiv preprint arXiv:2307.08698, 2023

    2023 arXiv

  26. [34]

    Instaflow: One step is enough for high-quality diffusion-based text-to-image generation

    Xingchao Liu, Xiwen Zhang, Jianzhu Ma, Jian Peng, and qiang liu. Instaflow: One step is enough for high-quality diffusion-based text-to-image generation. In ICLR, 2024

    2024

  27. [35]

    Yaron Lipman, Ricky T. Q. Chen, Heli Ben-Hamu, Maximilian Nickel, and Matthew Le. Flow matching for generative modeling. In ICLR, 2023

    2023

  28. [36]

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

    Junsong Chen, Jincheng YU, Chongjian GE, Lewei Yao, Enze Xie, Zhongdao Wang, James Kwok, Ping Luo, Huchuan Lu, and Zhenguo Li. Pixart-$\alpha$: Fast training of diffusion transformer for photorealistic text-to-image synthesis. In ICLR, 2024

    2024

  29. [37]

    Pixart- \sigma: Weak-to-strong training of diffusion transformer for 4k text-to-image generation

    Junsong Chen, Chongjian Ge, Enze Xie, Yue Wu, Lewei Yao, Xiaozhe Ren, Zhongdao Wang, Ping Luo, Huchuan Lu, and Zhenguo Li. Pixart- \sigma: Weak-to-strong training of diffusion transformer for 4k text-to-image generation. arXiv preprint arXiv:2403.04692, 2024

    2024 arXiv

  30. [38]

    Playground v3: Improving text-to-image alignment with deep-fusion large language models

    Bingchen Liu, Ehsan Akhgari, Alexander Visheratin, Aleks Kamko, Linmiao Xu, Shivam Shrirao, Joao Souza, Suhail Doshi, and Daiqing Li. Playground v3: Improving text-to-image alignment with deep-fusion large language models. arXiv preprint arXiv:2409.10695, 2024

    2024 arXiv

  31. [39]

    Lumina-next : Making lumina-t2x stronger and faster with next-dit

    Le Zhuo, Ruoyi Du, Han Xiao, Yangguang Li, Dongyang Liu, Rongjie Huang, Wenze Liu, Xiangyang Zhu, Fu-Yun Wang, Zhanyu Ma, Xu Luo, Zehan Wang, Kaipeng Zhang, Lirui Zhao, Si Liu, Xiangyu Yue, Wanli Ouyang, Yu Qiao, Hongsheng Li, and Peng Gao. Lumina-next : Making lumina-t2x stro...

    2024

  32. [40]

    Lumina-video: Efficient and flexible video generation with multi-scale Next-DiT

    Dongyang Liu, Shicheng Li, Yutong Liu, Zhen Li, Kai Wang, Xinyue Li, Qi Qin, Yufei Liu, Yi Xin, Zhongyu Li, Bin Fu, Chenyang Si, Yuewen Cao, Conghui He, Ziwei Liu, Yu Qiao, Qibin Hou, Hongsheng Li, and Peng Gao. Lumina-video: Efficient and flexible video generation with multi-...

    2025 arXiv

  33. [41]

    Sana 1.5: Efficient scaling of training-time and inference-time compute in linear diffusion transformer

    Enze Xie, Junsong Chen, Yuyang Zhao, Jincheng Yu, Ligeng Zhu, Chengyue Wu, Yujun Lin, Zhekai Zhang, Muyang Li, Junyu Chen, Han Cai, Bingchen Liu, Daquan Zhou, and Song Han. Sana 1.5: Efficient scaling of training-time and inference-time compute in linear diffusion transformer....

    2025 arXiv

  34. [42]

    Scaling vision transformers

    Xiaohua Zhai, Alexander Kolesnikov, Neil Houlsby, and Lucas Beyer. Scaling vision transformers. In CVPR, 2022

    2022

  35. [43]

    Scaling laws for diffusion transformers

    Zhengyang Liang, Hao He, Ceyuan Yang, and Bo Dai. Scaling laws for diffusion transformers. arXiv preprint arXiv:2410.08184, 2024

    2024

  36. [44]

    Progressive distillation for fast sampling of diffusion models

    Tim Salimans and Jonathan Ho. Progressive distillation for fast sampling of diffusion models. In ICLR, 2022. 11

    2022

  37. [45]

    Adversarial diffusion distillation

    Axel Sauer, Dominik Lorenz, Andreas Blattmann, and Robin Rombach. Adversarial diffusion distillation. In ECCV, 2024

    2024

  38. [46]

    Latent consistency models: Synthesizing high-resolution images with few-step inference

    Simian Luo, Yiqin Tan, Longbo Huang, Jian Li, and Hang Zhao. Latent consistency models: Synthesizing high-resolution images with few-step inference. arXiv preprint arXiv:2310.04378, 2023

    2023 arXiv

  39. [47]

    Trajectory consistency distillation

    Jianbin Zheng, Minghui Hu, Zhongyi Fan, Chaoyue Wang, Changxing Ding, Dacheng Tao, and Tat-Jen Cham. Trajectory consistency distillation. arXiv preprint arXiv:2402.19159, 2024

    2024 arXiv

  40. [48]

    Hyper-SD: Trajectory segmented consistency model for efficient image synthesis

    Yuxi Ren, Xin Xia, Yanzuo Lu, Jiacheng Zhang, Jie Wu, Pan Xie, XING WANG, and Xuefeng Xiao. Hyper-SD: Trajectory segmented consistency model for efficient image synthesis. In NeurIPS, 2024

    2024

  41. [49]

    One-step diffusion with distribution matching distillation

    Tianwei Yin, Michaël Gharbi, Richard Zhang, Eli Shechtman, Frédo Durand, William T Freeman, and Taesung Park. One-step diffusion with distribution matching distillation. In CVPR, 2024

    2024

  42. [50]

    Improved distribution matching distillation for fast image synthesis

    Tianwei Yin, Michaël Gharbi, Taesung Park, Richard Zhang, Eli Shechtman, Fredo Durand, and William T Freeman. Improved distribution matching distillation for fast image synthesis. In NeurIPS, 2024

    2024

  43. [51]

    Swiftbrush v2: Make your one-step diffusion model better than its teacher

    Trung Dao, Thuan Hoang Nguyen, Thanh Le, Duc Vu, Khoi Nguyen, Cuong Pham, and Anh Tran. Swiftbrush v2: Make your one-step diffusion model better than its teacher. In ECCV, 2024

    2024

  44. [52]

    Fast high-resolution image synthesis with latent adversarial diffusion distillation

    Axel Sauer, Frederic Boesel, Tim Dockhorn, Andreas Blattmann, Patrick Esser, and Robin Rom- bach. Fast high-resolution image synthesis with latent adversarial diffusion distillation. arXiv preprint arXiv:2403.12015, 2024

    2024 arXiv

  45. [53]

    Sana-sprint: One-step diffusion with continuous-time consistency distillation

    Junsong Chen, Shuchen Xue, Yuyang Zhao, Jincheng Yu, Sayak Paul, Junyu Chen, Han Cai, Enze Xie, and Song Han. Sana-sprint: One-step diffusion with continuous-time consistency distillation. arXiv preprint arXiv: 2503.09641, 2025

    2025

  46. [54]

    Diffusion adversarial post-training for one-step video generation

    Shanchuan Lin, Xin Xia, Yuxi Ren, Ceyuan Yang, Xuefeng Xiao, and Lu Jiang. Diffusion adversarial post-training for one-step video generation. arXiv preprint arXiv:2501.08316, 2025

    2025

  47. [55]

    Analysis of classifier-free guidance weight schedulers

    WANG Xi, Nicolas Dufour, Nefeli Andreou, CANI Marie-Paule, Victoria Fernandez Abrevaya, David Picard, and Vicky Kalogeiton. Analysis of classifier-free guidance weight schedulers. TMLR, 2024

    2024

  48. [56]

    CFG++: Manifold- constrained classifier free guidance for diffusion models

    Hyungjin Chung, Jeongsol Kim, Geon Yeong Park, Hyelin Nam, and Jong Chul Ye. CFG++: Manifold- constrained classifier free guidance for diffusion models. In ICLR, 2025

    2025

  49. [57]

    Yeh, and Ziwei Liu

    Weichen Fan, Amber Yijia Zheng, Raymond A. Yeh, and Ziwei Liu. Cfg-zero*: Improved classifier-free guidance for flow matching models. arXiv preprint arXiv:2503.18886, 2025

    2025 arXiv

  50. [58]

    Improving sample quality of diffusion models using self-attention guidance

    Susung Hong, Gyuseong Lee, Wooseok Jang, and Seungryong Kim. Improving sample quality of diffusion models using self-attention guidance. In ICCV, 2023

    2023

  51. [59]

    Self-rectifying diffusion sampling with perturbed-attention guidance

    Donghoon Ahn, Hyoungwon Cho, Jaewon Min, Wooseok Jang, Jungwoo Kim, SeonHwa Kim, Hyun Hee Park, Kyong Hwan Jin, and Seungryong Kim. Self-rectifying diffusion sampling with perturbed-attention guidance. In ECCV, 2024

    2024

  52. [60]

    Microsoft coco: Common objects in context

    Tsung-Yi Lin, Michael Maire, Serge Belongie, James Hays, Pietro Perona, Deva Ramanan, Piotr Dollár, and C Lawrence Zitnick. Microsoft coco: Common objects in context. In ECCV, 2014

    2014

  53. [61]

    Sdxl: Improving latent diffusion models for high-resolution image synthesis

    Dustin Podell, Zion English, Kyle Lacey, Andreas Blattmann, Tim Dockhorn, Jonas Müller, Joe Penna, and Robin Rombach. Sdxl: Improving latent diffusion models for high-resolution image synthesis. arXiv preprint arXiv:2307.01952, 2023

    2023 arXiv

  54. [62]

    SANA: Efficient high-resolution text-to-image synthesis with linear diffusion transformers

    Enze Xie, Junsong Chen, Junyu Chen, Han Cai, Haotian Tang, Yujun Lin, Zhekai Zhang, Muyang Li, Ligeng Zhu, Yao Lu, and Song Han. SANA: Efficient high-resolution text-to-image synthesis with linear diffusion transformers. In ICLR, 2025

    2025

  55. [63]

    Flux.1-turbo-alpha

    Alimama Smart Creative and AI Application Team. Flux.1-turbo-alpha. https://huggingface.co/ alimama-creative/FLUX.1-Turbo-Alpha , 2024

    2024

  56. [64]

    Playground v2.5: Three insights towards enhancing aesthetic quality in text-to-image generation

    Daiqing Li, Aleks Kamko, Ehsan Akhgari, Ali Sabet, Linmiao Xu, and Suhail Doshi. Playground v2.5: Three insights towards enhancing aesthetic quality in text-to-image generation. arXiv preprint arXiv:2402.17245, 2024

    2024 arXiv

  57. [65]

    NAG” and the small words “Normalized Attention Guidance: Universal Negative Guidance for Diffusion Models

    Yuval Kirstain, Adam Polyak, Uriel Singer, Shahbuland Matiana, Joe Penna, and Omer Levy. Pick-a-pic: An open dataset of user preferences for text-to-image generation. In NeurIPS, 2023. 12 Table of Contents 1 Introduction 2 2 Related Works 3 3 Background 3 3.1 Text-to-Image Dif...

    2023

  58. [66]

    An anthropomorphic cat thoughtfully paints an oil self-portrait on canvas, capturing its likeness with delicate brushstrokes inside a warmly lit, artistically cluttered studio

  59. [67]

    The fox is made of polygons

    An origami fox running in the forest. The fox is made of polygons. Speed and passion. Realistic. Figure 21 and Figure 22: First row:

  60. [68]

    A cabinet in which all the planets of the solar system are collected

  61. [69]

    Cyborg bird

    a futuristic interpretation of a dodo bird. Cyborg bird. Amazing colorful. Artstation, hyperrealistic

  62. [70]

    a happy female wizard surrounded by pieces of paper flying in the air around her

  63. [71]

    a recruitment consultant, sitting before a screen full of analysis diagram, carrying mobile device, fuji film style, like moss in wandering earth Second row:

  64. [72]

    Large birthday cake for a cardiothoracic surgeon

  65. [73]

    an anthropomorphic white rabbit, male wizard face, dressed in black and white, fine art, award-winning, intricate, elegant, sharp focus, cinematic lighting, highly detailed, digital painting, 8 k concept art, art by guweiz and z. w. gu, masterpiece, trending on artstation, 8 k

  66. [74]

    an owl transforms into an eagle

  67. [75]

    a photorealistic dragon pup Third row:

  68. [76]

    in a room a MGb car smashing through hole in the wall ,sparks dust rubble bricks ,studio lighting,white walls, mg logo

  69. [77]

    a werewolf reading a book

  70. [78]

    Black and white 1905 year portrait of futuristic professional photographer with camera in hand sadly seating deep in a dark pit covered by splash of dust

    1905

  71. [79]

    a panda riding a motorcycle Fourth row:

  72. [80]

    a wide angle photo of roman soldiers in front of courtyard roman buildings,technicolor film ,roman soldier in foreground masculine features nose helmet and silver sword ,eyes,clear sky, arches grass steps field panorama,Canaletto,stone floor,vanishing point,ben-hur flags , a d...

  73. [81]

    cuddly stuffed dinosaur talking to a microphone

  74. [82]

    blue apple, red banana

  75. [83]

    Anime cute little girl Fifth row:

  76. [84]

    a lonely man inside a old bucolic house surrounded by cats by Richard Billingham

  77. [85]

    full body space suit with boots, futuristic, character design, cinematic lightning, epic fantasy, hyper realistic, detail 8k 26

  78. [86]

    a young danish traveller standing at an immigration counter in ho chi minh city

  79. [87]

    Amigurumi figure of a little pig wearing a red sweater, professional photography, close up, vintage, 8k, product photo J.2 Prompts for User Study Table 10: Positive-negative prompt pairs used in the user study comparing NAG and NASA. Positive prompt Negative prompt A photo of ...

Pith tools

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