Pith. sign in

REVIEW 5 major objections 5 minor 6 cited by

NitroFusion: High-Fidelity Single-Step Diffusion through Dynamic Adversarial Training

T0 review · 5 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read NitroFusion claims that a continuously refreshing pool of specialized discriminator heads lets a single-step diffusion model match or beat multi-step teachers on fine detail and global coherence.

desk verdict Genuinely new adversarial-distillation mechanism with flexible 1–4 step deployment, but its own Table 1 contradicts the abstract's claim of significant outperformance. read the letter →

arxiv 2412.02030 v2 pith:T5URWV2C submitted 2024-12-02 cs.CV

Dar-Yen Chen , Hmrishav Bandyopadhyay , Kai Zou , Yi-Zhe Song This is my paper · ORCID
classification cs.CV
keywords single-stepdiffusionadversarialdistillationdiscriminatorpooltext-to-imagegenerationGANtrainingdynamicrefreshmulti-scalediscrimination
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

NitroFusion aims to close the quality gap between one-step and multi-step text-to-image diffusion. Instead of a single discriminator, it maintains a pool of hundreds of lightweight discriminator heads on top of a frozen teacher encoder, each specialized for a noise level and image scale, and it randomly re-initializes about 1% of these heads every iteration. The paper argues that this 'panel of critics' keeps feedback diverse and stable, allowing a student model to compress the full denoising trajectory into one step without losing fine detail or global consistency. If the claim is correct, real-time image generation can approach multi-step teacher quality at a fraction of the inference cost, and the same model can be run for 1 to 4 steps to trade quality against speed.

What carries the argument

The central object is the dynamic discriminator pool: a set of 480 lightweight heads mounted on the teacher's frozen UNet encoder, with 160 heads in each of three compartments (global conditional, local conditional, local unconditional). Each head processes features at a specific noise level t*, and the pool is sampled stochastically each training iteration, with a periodic refresh that re-initializes roughly 1% of the heads. This pool feeds the adversarial loss that trains a one-step student; the paper credits the pool's size, compartmentalization, and refresh with providing the diverse, stable feedback that lets the student avoid the blur and artifacts typical of one-step distillation.

What would settle it

Re-run the full NitroFusion recipe on the same teacher with the pool refresh rate set to 0%, 1%, and 5% and evaluate Patch Teacher FID, Aesthetic Score, and ImageReward; the paper predicts roughly 1% is optimal, so a flat or inverted curve with respect to refresh rate would falsify the refresh mechanism's claimed contribution.

Watch

Extended reading notes

Core claim

The paper's central claim is that high-fidelity one-step diffusion distillation is achievable with a dynamic adversarial framework built from a large pool of specialized discriminator heads. The pool is compartmentalized by discriminator timestep and by task—global conditional, local conditional, and local unconditional—so that different heads critique global coherence, local texture, and prompt alignment. A subset of heads is sampled each iteration to compute the adversarial loss, and roughly 1% of the pool is randomly re-initialized to prevent discriminator overfitting. The authors report that the resulting one-step models, NitroSD-Realism and NitroSD-Vibrant, outperform existing one-step baselines on aesthetic score and ImageReward, and that user studies prefer them over multi-step teachers and even 25-step SDXL. They also show that the same weights can produce 2-, 3-, and 4-step outputs that refine quality, and that the framework can be adapted to custom SDXL checkpoints through weight differences alone.

Load-bearing premise

The load-bearing assumption is that randomly re-initializing about 1% of the discriminator heads each iteration preserves the pool's collective knowledge and prevents overfitting, so that the dynamic feedback stays diverse; if the refresh instead discards useful learned signal or just adds noise, the claimed advantage of the dynamic pool collapses.

Editorial extensions

If this is right

  • Single-step text-to-image generation can reach quality competitive with 4- to 25-step baselines, making real-time interactive generation practical without a visible fidelity penalty.
  • The same model can serve 1-4 denoising steps, letting applications choose a latency/quality trade-off at inference time rather than maintaining separate checkpoints.
  • Distilling through a dynamic discriminator pool can transfer a teacher's stylistic identity (photorealism versus vibrant color) to a one-step student.
  • Because the method is trained purely on synthetic teacher samples, it can be adapted to specialized styles or domains without paired real-image data, as shown for anime and oil-painting checkpoints.

Reading between the lines

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

  • The paper's own ablations show that removing the pool refresh raises Aesthetic Score and ImageReward slightly, which suggests the load-bearing component may be the size and compartmentalization of the pool rather than the refresh rate; sweeping refresh rates from 0% to 5% on a single teacher would separate these effects.
  • The 'panel of critics' analogy implies an ensemble-diversity mechanism; a direct test is to measure the agreement rate among sampled heads during training—if refreshed heads rapidly converge to the predictions of retained heads, then the refresh injects no genuine diversity.
  • Since the same weights are used in the multi-step regime, a natural testable extension is whether classifier-free guidance, which the paper notes is not supported, can be re-enabled for 2-4 step generation to improve prompt alignment without retraining.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 5 minor

Summary. The paper introduces NitroFusion, a single-step diffusion distillation framework built on a dynamic adversarial training scheme. The method maintains a large pool of lightweight discriminator heads attached to a frozen UNet encoder, compartmentalizes them by noise level and by global/local and conditional/unconditional objectives, and randomly refreshes about 1% of the heads each iteration. The student generator is trained with an adversarial loss only, using a multi-step teacher as the source of real samples, and the same weights support 1-4 step inference through a bottom-up refinement procedure. The authors train two variants, NitroSD-Realism (DMD2 teacher) and NitroSD-Vibrant (Hyper-SDXL teacher), and evaluate on COCO-5K with CLIP, FID, Aesthetic Score, and ImageReward, plus a user study. The abstract claims that NitroFusion significantly outperforms existing single-step methods across multiple evaluation metrics.

Significance. If the claims were fully supported, the paper would make a useful contribution to one-step diffusion distillation: a large, dynamic pool of specialized discriminator heads is a plausible way to provide richer adversarial feedback, and the ability to refine from 1 to 4 steps with the same weights addresses a practical deployment need. The evaluation design has strengths: the main quantitative comparisons use external benchmarks rather than the teacher used for training, and the user study collects a substantial number of votes (2,884 votes from 170 participants in the supplementary material). The paper also demonstrates adaptation to custom SDXL checkpoints through weight differences, which is a useful practical result. However, the central quantitative claim of broad outperformance is only partially supported by the paper's own Table 1, and the ablation study in Supplementary Table 2 does not consistently support the role of the proposed components. These issues need to be resolved before the paper can be recommended for publication.

major comments (5)
  1. [Section 4.3, Table 1] The abstract's central claim that NitroFusion 'significantly outperforms existing single-step methods across multiple evaluation metrics' is not supported by Table 1. At one step, NitroSD-Vibrant is numerically worse than Hyper-SDXL on all four reported metrics: CLIP 0.314 vs 0.317, FID 38.49 vs 36.77, Aesthetic 5.92 vs 6.00, and ImageReward 0.991 vs 1.169. NitroSD-Realism is worse than DMD2 on FID (25.61 vs 23.91), tied on CLIP (0.320 vs 0.320), and better only on Aesthetic (5.56 vs 5.47) and ImageReward (0.856 vs 0.825). The conclusion that 'our model outperforms state-of-the-art low-step and one-step baselines in both qualitative and quantitative analysis' needs to be revised to a more limited claim, or additional evidence is needed to support the broad statement.
  2. [Supplementary Table 2, Section 4.5] The ablation study does not support the claimed benefit of the pool refresh mechanism. Removing pool refresh improves Aesthetic Score from 5.87 to 5.98 and ImageReward from 1.020 to 1.054, while only slightly worsening Patch Teacher FID (18.78 vs 18.70). This contradicts the qualitative statement in Section 4.5 that without Pool Refresh 'artifacts persist and sharpness is lost'. The authors should either provide an explanation for this discrepancy or temper the claim about the necessity of refresh.
  3. [Section 4.3 and Supplementary Table 2] No error bars, confidence intervals, or significance tests are reported for any quantitative metric. The differences that support the paper's claims are small (e.g., Aesthetic 5.56 vs 5.47, ImageReward 0.856 vs 0.825), and without variance estimates it is impossible to judge whether they are meaningful. The word 'significantly' in the abstract is therefore not justified by the reported statistics.
  4. [Section 4.2, Figure 5] The user study reports preference percentages without variance, confidence intervals, or statistical significance tests. For example, a 53.2% preference for NitroSD-Realism over SDXL-Turbo in one comparison is close to chance and may not be significant even with the reported 2,884 total votes. The user study is a preference measure; it does not constitute the 'multiple evaluation metrics' promised in the abstract, and its results need to be reported with appropriate uncertainty.
  5. [Supplementary Section B] The Patch Teacher FID metric is defined as the FID between student and teacher samples, and the student is explicitly trained to match the teacher. This makes Patch Teacher FID a distillation-fidelity measure rather than an independent quality metric. The statement that 'removing each component causes varying levels of degradation in Patch Teacher FID' is partially expected from the training objective and cannot by itself establish that the components improve real-image fidelity. The metric should be framed as a diagnostic of teacher matching, not as a general quality metric.
minor comments (5)
  1. [Equations (3)-(6)] The notation for the discriminator output is inconsistent: D(ˆx0) in Equations (3)-(4) is later written as a sum over heads in Equations (5)-(6). Please define whether D denotes a single head, a sampled subset, or the aggregated multi-head output, and keep the notation uniform.
  2. [Algorithm 1, line 20] The line 'Prefresh ∼ N(0, I)' is not a clear description of re-initializing discriminator heads. Since the heads have structured weights, please specify the initialization distribution and how the refreshed heads are inserted into the pool.
  3. [Section 2.2] The related work mentions 'SDXL-Lighting' but the correct name in the rest of the paper is 'SDXL-Lightning'. Please fix this typo.
  4. [Figures 4 and 5] The captions of Figures 4 and 5 are hard to parse because 'Teacher' and 'Student' labels are placed above columns without clarifying which row corresponds to which model, and the model names are listed below. Please restructure the captions to unambiguously identify each panel.
  5. [Section 4.6] The zero-shot adaptation to custom checkpoints is interesting, but the description of applying 'the weight difference between NitroFusion and SDXL' is vague. Please specify how the weight difference is computed and applied, and whether any normalization is used.

Circularity Check

1 steps flagged · score 2.0 of 10

Main claim is externally benchmarked; only the supplementary Patch Teacher FID metric is circular by construction.

  1. self definitional [Supplementary Material, Section B (Additional Ablation Study), paragraph introducing Patch Teacher FID and Table 2]
    "In particular, we introduce the Patch Teacher FID metric, which measures the FID score between 299 × 299 center-cropped patches from student and teacher samples [23], assessing how well high-resolution details are preserved. This metric serves as a critical index for evaluating the effectiveness of GAN training, as it emphasizes the generator’s ability to represent fine-grained features and maintain fidelity to the teacher model."

    The student is trained by Eqs. (5)-(6) to make its one-step output distribution match the teacher's output distribution through an adversarial loss pitting real teacher samples against fake student samples under the teacher's frozen UNet encoder. Patch Teacher FID is exactly a distributional distance between student patches and teacher patches, so it measures the training objective itself rather than an independent quality quantity. Using it as 'a critical index for evaluating the effectiveness of GAN training' thus makes the ablation conclusion self-referential: showing that removing components worsens Patch Teacher FID only shows that those components help optimize teacher-matching, not that they independently improve generation quality.

full rationale

The main derivation chain is not circular. NitroFusion is an adversarial distillation method: a one-step student is trained with a dynamic pool of discriminator heads against a frozen teacher UNet encoder, and the central claim of high-fidelity one-step generation is evaluated on external benchmarks—COCO-5K FID, CLIP score, Aesthetic Score, ImageReward, and a human preference study—rather than on the teacher's own outputs. The teacher is used as a training signal, but the external metrics break the loop. No load-bearing self-citations appear: the technical components build on prior work by other groups (SDXL-Lightning, DMD2, Hyper-SDXL, StyleGAN-XL/T), and no uniqueness theorem or ansatz is imported from the present authors' own prior work. The one genuinely self-referential element is the supplementary Patch Teacher FID metric, which measures exactly the student-teacher distribution match that the adversarial loss optimizes; because it is used to attribute ablations, it is mildly circular, but it is not the paper's main evidence. Separately, there is a substantive evidence-quality concern—Table 1 does not support the abstract's claim of 'significantly outperforming' across all metrics, and the pool-refresh ablation is mixed—but that is a correctness or reporting concern, not circularity.

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

No new physical entities are introduced. The framework's novel objects are the discriminator pool and refresh mechanism, which are engineering constructs rather than invented entities with independent evidence.

free parameters (4)
  • Discriminator pool size = 480 heads (160 per task type)
    Chosen by hand; no sensitivity analysis provided.
  • Pool refresh rate = 1% per iteration
    Set to 1% with no ablation across rates.
  • Discriminator timesteps = {10, 250, 500, 750} (shifted variants in Supp. A)
    Selected based on prior work [23]; no study of alternative sets.
  • Global/local head ratio = 1:2 (global conditional : local conditional : local unconditional)
    Chosen by hand; no analysis of other ratios.
assumptions (3)
  • domain assumption Synthetic teacher-generated samples are sufficient for distillation training without real paired prompt-image data (following [42]).
    The paper trains only on 1M teacher-generated samples; if teacher outputs have systematic flaws, the student will inherit them.
  • domain assumption The frozen UNet encoder features at various noise levels and spatial scales provide informative, stable discriminator signals.
    The method attaches all heads to a frozen SDXL encoder; the paper does not compare with trainable backbones or other feature extractors.
  • domain assumption Adversarial loss alone can align student and teacher distributions without score-matching or flow-matching objectives.
    The training uses only adversarial losses (Eq. 5-6); this is a strong assumption given prior distillation methods combine objectives.

how reviews work

0 comments
Cite this review

Pith. "Pith review of NitroFusion: High-Fidelity Single-Step Diffusion through Dynamic Adversarial Training." pith.science (2026). https://pith.science/paper/T5URWV2C

@misc{pith2026241202030,
  author       = {Pith},
  title        = {Pith review of: NitroFusion: High-Fidelity Single-Step Diffusion through Dynamic Adversarial Training},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/T5URWV2C}},
  note         = {Machine review of arXiv:2412.02030}
}
read the original abstract

We introduce NitroFusion, a fundamentally different approach to single-step diffusion that achieves high-quality generation through a dynamic adversarial framework. While one-step methods offer dramatic speed advantages, they typically suffer from quality degradation compared to their multi-step counterparts. Just as a panel of art critics provides comprehensive feedback by specializing in different aspects like composition, color, and technique, our approach maintains a large pool of specialized discriminator heads that collectively guide the generation process. Each discriminator group develops expertise in specific quality aspects at different noise levels, providing diverse feedback that enables high-fidelity one-step generation. Our framework combines: (i) a dynamic discriminator pool with specialized discriminator groups to improve generation quality, (ii) strategic refresh mechanisms to prevent discriminator overfitting, and (iii) global-local discriminator heads for multi-scale quality assessment, and unconditional/conditional training for balanced generation. Additionally, our framework uniquely supports flexible deployment through bottom-up refinement, allowing users to dynamically choose between 1-4 denoising steps with the same model for direct quality-speed trade-offs. Through comprehensive experiments, we demonstrate that NitroFusion significantly outperforms existing single-step methods across multiple evaluation metrics, particularly excelling in preserving fine details and global consistency.

Figures

Figures reproduced from arXiv: 2412.02030 by the authors.

Figure 1
Figure 1. Our one-step diffusion pipeline generates vibrant and photorealistic images with exceptional detail in a single inference step, [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Our method distils a multi-step teacher model into an efficient one-step student generator. The Dynamic Adversarial Framework [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Our discriminator employs a frozen UNet backbone with [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (9 more)
Figure 4
Figure 4. Figure 4: Visual comparison of our models (NitroSD-Realism and NitroSD-Vibrant) against multi-step SDXL [ [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: User preferences study with other baseline models. [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: Visual comparison of our models (NitroSD-Realism and NitroSD-Vibrant) with other approaches across multiple steps, high [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 8
Figure 8. Figure 8: Results from applying NitroSD-Realism to anime [ [PITH_FULL_IMAGE:figures/full_fig_p008_8.png]
Figure 7
Figure 7. Figure 7: Qualitative study of ablative configurations [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]
Figure 9
Figure 9. Figure 9: 1- to 4-step refinement process of our NitroSD-Realism and -Vibrant, illustrating the progressive enhancement of image quality [PITH_FULL_IMAGE:figures/full_fig_p011_9.png]
Figure 10
Figure 10. Figure 10: Additional visual comparison with state-of-the-art approaches. [PITH_FULL_IMAGE:figures/full_fig_p013_10.png]
Figure 11
Figure 11. Figure 11: Additional single-step samples from NitroSD-Realism. [PITH_FULL_IMAGE:figures/full_fig_p014_11.png]
Figure 12
Figure 12. Figure 12: Additional single-step samples from NitroSD-Vibrant. [PITH_FULL_IMAGE:figures/full_fig_p015_12.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 6 Pith papers

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

  1. Dual-Expert Consistency Model for Efficient and High-Quality Video Generation

    cs.CV 2025-06 conditional novelty 6.0 of 10

    By training a semantic expert and a LoRA-based detail expert, DCM reaches nearly teacher-level VBench scores with 4-step video sampling on HunyuanVideo and CogVideoX.

  2. Integrating Spatial and Semantic Embeddings for Stereo Sound Event Localization in Videos

    eess.AS 2025-09 conditional novelty 5.0 of 10

    A Cross-Modal Conformer that fuses CLAP audio and OWL-ViT visual embeddings with a CNN-Conformer SELD backbone, trained on large synthetic data, ranks second in DCASE 2025 Task 3 Track B.

  3. FVGen: Accelerating Novel-View Synthesis with Adversarial Video Diffusion Distillation

    cs.CV 2025-08 unverdicted novelty 5.0 of 10

    FVGen uses GAN-based adversarial distillation and softened reverse KL divergence to compress a video diffusion teacher for novel-view synthesis into a four-step student with comparable quality.

  4. Spatial and Semantic Embedding Integration for Stereo Sound Event Localization and Detection in Regular Videos

    eess.AS 2025-07 conditional novelty 5.0 of 10

    Fusing frozen CLAP and OWL-ViT embeddings via a Cross-Modal Conformer, plus autocorrelation-based features, improves stereo SELD over DCASE 2025 baselines.

  5. Normalized Attention Guidance: Universal Negative Guidance for Diffusion Models

    cs.CV 2025-05 conditional novelty 5.0 of 10

    Normalized Attention Guidance (NAG) stabilizes attention-space extrapolation with L1 normalization and refinement, restoring negative prompting in few-step diffusion models across architectures and modalities.

  6. Few-Step Diffusion via Score identity Distillation

    cs.CV 2025-05 conditional novelty 5.0 of 10

    A few-step, largely data-free extension of Score identity Distillation reaches state-of-the-art FID and CLIP scores on SDXL at 1024x1024 with one or four generation steps.

Reference graph

Works this paper leans on

58 extracted references · 31 canonical work pages · cited by 6 Pith papers

  1. [1]

    https : / / github

    Clip+mlp aesthetic score predictor. https : / / github . com / christophschuhmann / improved - aesthetic-predictor, 2022. 7

    work page 2022

  2. [2]

    https : / / github

    Low-rank adaptation for fast text-to-image diffusion fine- tuning. https : / / github . com / cloneofsimo / lora, 2022. 8

    work page 2022

  3. [3]

    https://civitai.com/models/ 260267/animagine-xl-v31, 2024

    Animagine xl v3.1. https://civitai.com/models/ 260267/animagine-xl-v31, 2024. 8

    work page 2024

  4. [4]

    https://civitai.com/, 2024

    Civitai. https://civitai.com/, 2024. 8

    work page 2024

  5. [5]

    https://civitai.com/models/240154/ painters - checkpoint - oil - paint - oil - painting-art-style, 2024

    Painter’s checkpoint (oil paint / oil painting art style) v1.1. https://civitai.com/models/240154/ painters - checkpoint - oil - paint - oil - painting-art-style, 2024. 8

    work page 2024

  6. [6]

    Multi- Objective Training of Generative Adversarial Networks with Multiple Discriminators

    Isabela Albuquerque, Joao Monteiro, Thang Doan, Brean- dan Considine, Tiago Falk, and Ioannis Mitliagkas. Multi- Objective Training of Generative Adversarial Networks with Multiple Discriminators. In ICML, 2019. 4

    work page 2019

  7. [7]

    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. 11

    arXiv 2024

  8. [8]

    MCL-GAN: Generative adversarial networks with multiple specialized discrimina- tors

    Jinyoung Choi and Bohyung Han. MCL-GAN: Generative adversarial networks with multiple specialized discrimina- tors. In NeurIPS, 2022. 2, 3, 4

    work page 2022

Show all 58 references
  1. [9]

    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. 3, 4

    2024

  2. [10]

    Diffusion models beat gans on image synthesis

    Prafulla Dhariwal and Alexander Nichol. Diffusion models beat gans on image synthesis. In NeurIPS, 2021. 12

    2021

  3. [11]

    An image is worth 16x16 words: Transformers for image recognition at scale

    Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Syl- vain Gelly, Jakob Uszkoreit, and Neil Houlsby. An image is worth 16x16 words: Transformers for image recognition ...

    2021

  4. [12]

    Gener- ative multi-adversarial networks

    Ishan Durugkar, Ian Gemp, and Sridhar Mahadevan. Gener- ative multi-adversarial networks. In ICLR, 2017. 2, 3

    2017

  5. [13]

    Generative Adversarial Nets

    Ian Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, and Yoshua Bengio. Generative Adversarial Nets. In NeurIPS,

  6. [14]

    Boot: Data-free distillation of denoising diffusion models with bootstrapping

    Jiatao Gu, Shuangfei Zhai, Yizhe Zhang, Lingjie Liu, and Joshua M Susskind. Boot: Data-free distillation of denoising diffusion models with bootstrapping. In ICMLW, 2023. 2, 3

    2023

  7. [15]

    Mul- tistep consistency models

    Jonathan Heek, Emiel Hoogeboom, and Tim Salimans. Mul- tistep consistency models. arXiv preprint arXiv:2403.06807,

    arXiv

  8. [16]

    Gaussian error linear units (gelus)

    Dan Hendrycks and Kevin Gimpel. Gaussian error linear units (gelus). arXiv preprint arXiv:1606.08415, 2023. 5

    2023 arXiv

  9. [17]

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

    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 equilib- rium. In NeurIPS, 2017. 7

    2017

  10. [18]

    Classifier-free diffusion guidance

    Jonathan Ho and Tim Salimans. Classifier-free diffusion guidance. In arXiv preprint arXiv:2207.12598, 2022. 12

    2022 arXiv

  11. [19]

    Denoising diffu- sion probabilistic models

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising diffu- sion probabilistic models. In NeurIPS, 2020. 3

    2020

  12. [20]

    LoRA: Low-rank adaptation of large language models

    Edward J Hu, yelong shen, Phillip Wallis, Zeyuan Allen- Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. LoRA: Low-rank adaptation of large language models. In ICLR, 2022. 8

    2022

  13. [21]

    Consistency trajectory mod- els: Learning probability flow ode trajectory of diffusion

    Dongjun Kim, Chieh-Hsin Lai, Wei-Hsiang Liao, Naoki Mu- rata, Yuhta Takida, Toshimitsu Uesaka, Yutong He, Yuki Mitsufuji, and Stefano Ermon. Consistency trajectory mod- els: Learning probability flow ode trajectory of diffusion. In ICLR, 2024. 2, 3

    2024

  14. [22]

    Pick-a-pic: An open dataset of user preferences for text-to-image generation

    Yuval Kirstain, Adam Polyak, Uriel Singer, Shahbuland Ma- tiana, Joe Penna, and Omer Levy. Pick-a-pic: An open dataset of user preferences for text-to-image generation. In NeurIPS, 2023. 6

    2023

  15. [23]

    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. 2, 3, 4, 5, 6, 7, 8, 11, 12

    2024 arXiv

  16. [24]

    Microsoft coco: Common objects in context

    Tsung-Yi Lin, Michael Maire, Serge Belongie, James Hays, Pietro Perona, Deva Ramanan, Piotr Doll´ar, and C Lawrence Zitnick. Microsoft coco: Common objects in context. In ECCV, 2014. 7

    2014

  17. [25]

    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. 2, 3

    2023

  18. [26]

    Decoupled weight decay regularization

    Ilya Loshchilov and Frank Hutter. Decoupled weight decay regularization. arXiv preprint arXiv:1711.05101, 2019. 5

    2019 arXiv

  19. [27]

    Knowledge distillation in iterative generative models for improved sampling speed

    Eric Luhman and Troy Luhman. Knowledge distillation in iterative generative models for improved sampling speed. arXiv preprint arXiv:2101.02388, 2021. 2, 3

    2021 arXiv

  20. [28]

    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. 3, 12

    2023 arXiv

  21. [29]

    On distillation of guided diffusion models

    Chenlin Meng, Robin Rombach, Ruiqi Gao, Diederik Kingma, Stefano Ermon, Jonathan Ho, and Tim Salimans. On distillation of guided diffusion models. In CVPR, 2023. 2, 3

    2023

  22. [30]

    Stabilizing gan training with multiple random projections

    Behnam Neyshabur, Srinadh Bhojanapalli, and Ayan Chakrabarti. Stabilizing gan training with multiple random projections. arXiv preprint arXiv:1705.07831, 2018. 2, 3

    2018 arXiv

  23. [31]

    Dual Dis- criminator Generative Adversarial Nets

    Tu Nguyen, Trung Le, Hung Vu, and Dinh Phung. Dual Dis- criminator Generative Adversarial Nets. In NeurIPS, 2017. 2, 3

    2017

  24. [32]

    Swiftbrush: One-step text-to-image diffusion model with variational score distilla- tion

    Thuan Hoang Nguyen and Anh Tran. Swiftbrush: One-step text-to-image diffusion model with variational score distilla- tion. In CVPR, 2024. 2, 3

    2024

  25. [33]

    Maxime Oquab, Timoth ´ee Darcet, Theo Moutakanni, Huy V . V o, Marc Szafraniec, Vasil Khalidov, Pierre Fernandez, Daniel Haziza, Francisco Massa, Alaaeldin El-Nouby, Rus- sell Howes, Po-Yao Huang, Hu Xu, Vasu Sharma, Shang- Wen Li, Wojciech Galuba, Mike Rabbat, Mido Assran, Ni...

    2023 arXiv

  26. [34]

    Sdxl: Improving latent diffusion mod- els for high-resolution image synthesis

    Dustin Podell, Zion English, Kyle Lacey, Andreas Blattmann, Tim Dockhorn, Jonas M ¨uller, Joe Penna, and Robin Rombach. Sdxl: Improving latent diffusion mod- els for high-resolution image synthesis. arXiv preprint arXiv:2307.01952, 2023. 6, 7, 8, 12

    2023 arXiv

  27. [35]

    Learn- ing transferable visual models from natural language super- vision

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learn- ing transferable visual models from natural language super- vision. In ICML, 2021. 7

    2021

  28. [36]

    Prajit Ramachandran, Barret Zoph, and Quoc V . Le. Searching for activation functions. arXiv preprint arXiv:1710.05941, 2017. 5

    2017 arXiv

  29. [37]

    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. arXiv preprint arXiv:2404.13686, 2024. 2, 3, 4, 6, 7, 8, 11, 12

    2024 arXiv

  30. [38]

    High-resolution image syn- thesis with latent diffusion models

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Bj¨orn Ommer. High-resolution image syn- thesis with latent diffusion models. In CVPR, 2022. 2, 4, 5

    2022

  31. [39]

    Progressive distillation for fast sampling of diffusion models

    Tim Salimans and Jonathan Ho. Progressive distillation for fast sampling of diffusion models. In ICLR, 2022. 2, 3

    2022

  32. [40]

    Stylegan- xl: Scaling stylegan to large diverse datasets

    Axel Sauer, Katja Schwarz, and Andreas Geiger. Stylegan- xl: Scaling stylegan to large diverse datasets. InSIGGRAPH,

  33. [41]

    StyleGAN-T: Unlocking the power of GANs for fast large-scale text-to-image synthesis

    Axel Sauer, Tero Karras, Samuli Laine, Andreas Geiger, and Timo Aila. StyleGAN-T: Unlocking the power of GANs for fast large-scale text-to-image synthesis. In ICML, 2023. 3

    2023

  34. [42]

    Adversarial diffusion distillation

    Axel Sauer, Dominik Lorenz, Andreas Blattmann, and Robin Rombach. Adversarial diffusion distillation. arXiv preprint arXiv:2311.17042, 2023. 2, 3, 4, 6, 7, 8, 12

    2023 arXiv

  35. [43]

    Fast high- resolution image synthesis with latent adversarial diffusion distillation

    Axel Sauer, Frederic Boesel, Tim Dockhorn, Andreas Blattmann, Patrick Esser, and Robin Rombach. Fast high- resolution image synthesis with latent adversarial diffusion distillation. arXiv preprint arXiv:2403.12015, 2024. 2, 3, 11

    2024 arXiv

  36. [44]

    Laion-5b: An open large-scale dataset for training next generation image-text models

    Christoph Schuhmann, Romain Beaumont, Richard Vencu, Cade Gordon, Ross Wightman, Mehdi Cherti, Theo Coombes, Aarush Katta, Clayton Mullis, Mitchell Worts- man, et al. Laion-5b: An open large-scale dataset for training next generation image-text models. In NeurIPS, 2022. 6

    2022

  37. [45]

    Score-based generative modeling through stochastic differential equa- tions

    Yang Song, Jascha Sohl-Dickstein, Diederik P Kingma, Ab- hishek Kumar, Stefano Ermon, and Ben Poole. Score-based generative modeling through stochastic differential equa- tions. In ICLR, 2021. 3

    2021

  38. [46]

    Consistency models

    Yang Song, Prafulla Dhariwal, Mark Chen, and Ilya Sutskever. Consistency models. In ICML, 2023. 2, 3

    2023

  39. [47]

    Diffusion-gan: Training gans with diffusion

    Zhendong Wang, Huangjie Zheng, Pengcheng He, Weizhu Chen, and Mingyuan Zhou. Diffusion-gan: Training gans with diffusion. In ICLR, 2023. 4

    2023

  40. [48]

    Group normalization

    Yuxin Wu and Kaiming He. Group normalization. In IJCV,

  41. [49]

    Accelerating image gen- eration with sub-path linear approximation model

    Chen Xu, Tianhui Song, Weixin Feng, Xubin Li, Tiezheng Ge, Bo Zheng, and Limin Wang. Accelerating image gen- eration with sub-path linear approximation model. arXiv preprint arXiv:2404.13903, 2024. 2, 3

    2024 arXiv

  42. [50]

    Imagere- ward: 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. Imagere- ward: Learning and evaluating human preferences for text- to-image generation. NeurIPS, 2024. 6, 7

    2024

  43. [51]

    Perflow: Piecewise rectified flow as universal plug-and-play accelerator

    Hanshu Yan, Xingchao Liu, Jiachun Pan, Jun Hao Liew, Qiang Liu, and Jiashi Feng. Perflow: Piecewise rectified flow as universal plug-and-play accelerator. arXiv preprint arXiv:2405.07510, 2024. 2, 3

    2024 arXiv

  44. [52]

    Im- proved distribution matching distillation for fast image syn- thesis

    Tianwei Yin, Micha ¨el Gharbi, Taesung Park, Richard Zhang, Eli Shechtman, Fredo Durand, and William T Freeman. Im- proved distribution matching distillation for fast image syn- thesis. In NeurIPS, 2024. 2, 3, 4, 6, 7, 8, 11, 12

    2024

  45. [53]

    One-step diffusion with distribution matching distillation

    Tianwei Yin, Micha ¨el Gharbi, Richard Zhang, Eli Shecht- man, Fr´edo Durand, William T Freeman, and Taesung Park. One-step diffusion with distribution matching distillation. In CVPR, 2024. 2, 3

    2024

  46. [54]

    Scaling autoregres- sive models for content-rich text-to-image generation

    Jiahui Yu, Yuanzhong Xu, Jing Yu Koh, Thang Luong, Gun- jan Baid, Zirui Wang, Vijay Vasudevan, Alexander Ku, Yin- fei Yang, Burcu Karagol Ayan, et al. Scaling autoregres- sive models for content-rich text-to-image generation. arXiv preprint arXiv:2206.10789, 2022. 11

    2022 arXiv

  47. [55]

    Unifl: Improve stable diffusion via unified feedback learning

    Jiacheng Zhang, Jie Wu, Yuxi Ren, Xin Xia, Huafeng Kuang, Pan Xie, Jiashi Li, Xuefeng Xiao, Weilin Huang, Min Zheng, Lean Fu, and Guanbin Li. Unifl: Improve stable diffusion via unified feedback learning. CoRR, 2024. 6

    2024

  48. [56]

    Hipa: Enabling one-step text- to-image diffusion models via high-frequency-promoting adaptation

    Yifan Zhang and Bryan Hooi. Hipa: Enabling one-step text- to-image diffusion models via high-frequency-promoting adaptation. arXiv preprint arXiv:2311.18158, 2023. 2, 3

    2023 arXiv

  49. [57]

    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. 2, 3

    2024 arXiv

  50. [58]

    Score identity distillation: Exponentially fast distillation of pretrained diffusion models for one-step generation

    Mingyuan Zhou, Huangjie Zheng, Zhendong Wang, Mingzhang Yin, and Hai Huang. Score identity distillation: Exponentially fast distillation of pretrained diffusion models for one-step generation. In ICML, 2024. 2, 3 10 NitroFusion: High-Fidelity Single-Step Diffusion through Dyna...

    2024

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.