Pith. sign in

REVIEW 4 major objections 6 minor 36 references

Implementing Adaptations for Vision AutoRegressive Model

T0 review · 4 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read Fine-tuned Vision AutoRegressive image models beat diffusion adaptation on small datasets, but private fine-tuning lags far behind.

desk verdict Useful first benchmark on VAR adaptation with reproducible code, but the paper's own Table 1 contradicts its headline claim on CUB-200-2011. read the letter →

arxiv 2507.11441 v2 pith:JESQV7RJ submitted 2025-07-15 cs.CV cs.LG

Kaif Shaikh , Franziska Boenisch , Adam Dziedzic This is my paper
classification cs.CVcs.LG
keywords visionautoregressivemodelimagegenerationfine-tuningparameter-efficientLoRALayerNormtuningdifferentialprivacyDP-SGD
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 asks whether Vision AutoRegressive (VAR) generators, a recently introduced alternative to diffusion models, can be adapted to narrow downstream tasks such as medical data generation. It implements full fine-tuning, LoRA, and LayerNorm tuning for VAR, and benchmarks them against the diffusion state-of-the-art adaptation method, DiffFit, on five image datasets. The central finding is that VAR adaptations converge in very few steps and that full fine-tuning reaches lower Frechet Inception Distance (FID) than DiffFit on the reported benchmarks, while LoRA nearly matches full fine-tuning with a small parameter budget. The paper then adds differentially private fine-tuning via DP-SGD and reports that the resulting models need very large privacy budgets and still produce poor images, so private adaptation of VAR remains an open problem.

What carries the argument

The object that carries the argument is VAR's next-scale prediction: instead of predicting tokens in raster-scan order, VAR predicts a sequence of 2D token grids from low to high resolution, trained with a direct cross-entropy token-prediction loss. The authors credit this objective with VAR's fast convergence, since the gradient signal is not diluted across noise levels the way a diffusion denoising loss is. The adaptations under study are full fine-tuning (all parameters updated), LoRA (low-rank adapters inserted into the attention, projection, and Adaptive LayerNorm modules), LayerNorm tuning (only Adaptive LayerNorm parameters), and, for privacy, DP-SGD with per-sample gradient clipping, Gaussian noise, and augmentation multiplicity; making these work required patching the model's buffers and its forward function so that per-sample gradient accounting becomes possible.

What would settle it

Rerun DiT-XL-2 with DiffFit under the paper's stated protocol, generating as many images per class as the test set and computing FID against the test split with the same implementation, and compare with the Table 1 VAR numbers; the CUB-200-2011 row, where the paper's own table shows DiffFit at 5.48 versus VAR FFT at 5.74 and 5.58, is the decisive cell. For the DP claim, run DP fine-tuning of VAR-d16 on Oxford Flowers with augmentation multiplicity beyond k equal to 128 and more training steps; if FID approached the non-private range of roughly 12 to 13, the claim that private adaptation founders would be contradicted.

Watch

Extended reading notes

Core claim

The paper's central discovery, as its authors state it, is that a Vision AutoRegressive generator can be adapted to small domain-specific image datasets more effectively than the diffusion model representing the current state of the art: full fine-tuning of VAR-d16 and VAR-d20 reaches lower FID than DiffFit on the benchmark datasets they report, while costing far fewer training steps, and LoRA nearly matches full fine-tuning with only a fraction of the parameters. A second, negative discovery is that differentially private fine-tuning built on DP-SGD does not transfer cleanly to VAR: on Oxford Flowers even with epsilon equal to 10 and augmentation multiplicity k equal to 128, LoRA reaches FID around 63, compared with non-private LoRA FID around 13, and convergence is slow. The authors present these results as evidence that autoregressive image models are a viable and efficient alternative to diffusion models for non-private adaptation, and that private adaptation of VAR still needs fundamentally better methods.

Load-bearing premise

The whole comparison against diffusion rests on the assumption that the published DiffFit FID numbers were produced under the same evaluation protocol as the new VAR numbers, because the diffusion baselines were not rerun in this paper.

Editorial extensions

If this is right

  • Non-private VAR adaptation is a usable alternative to diffusion fine-tuning: full fine-tuning gives the best FID, LoRA gives near-full-tuning quality with a small trainable parameter budget, and LayerNorm tuning is the cheapest adapter.
  • Because VAR converges in a few thousand steps, adaptation runs can be much shorter than diffusion fine-tuning, especially on small datasets.
  • The patched implementations released with the paper should let other researchers apply LoRA and DP-SGD to VAR-family autoregressive image models.
  • Differentially private adaptation of VAR is not yet practical: acceptable image quality appears to require implausibly large privacy budgets, so privacy-preserving uses remain open.
  • Using augmentation multiplicity, the paper's DP experiments can be extended toward better private utility, though the reported gains are modest and the compute cost scales directly with the number of views.

Reading between the lines

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

  • The paper's own Table 1 contains one row that the headline 'outperforms DiffFit across all models and datasets' does not obviously cover: on CUB-200-2011, DiffFit's FID of 5.48 is lower than VAR-d16 FFT's 5.74 and VAR-d20 FFT's 5.58, so a like-for-like reproduction of that cell is the first thing to check.
  • The buffer and forward-pass patches described in the appendix are transferable: any autoregressive image model with registered non-trainable buffers and a multi-argument forward method could use the same conversion to properties and a single concatenated input tensor to make DP-SGD-style per-sample accounting work.
  • A natural next step, not tested in the paper, is a noise schedule matched to next-scale token prediction, for instance clipping gradients at the scale level or adding token-level noise multiplicity analogous to what diffusion-model DP work calls noise multiplicity.
  • If the head-to-head comparison is reproduced under one unified FID protocol, LoRA would likely become the default non-private VAR adapter, since it nearly matches full fine-tuning while requiring far fewer trainable parameters.
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 implements and benchmarks several adaptation methods for the Vision AutoRegressive (VAR) image generation model: full fine-tuning (FFT), LoRA, and LayerNorm tuning, on five downstream datasets, and compares the resulting FID scores with the diffusion-model adaptation method DiffFit. It also implements differentially private (DP) fine-tuning of VAR using DP-SGD with augmentation multiplicity, evaluated on Oxford Flowers. The central claims are that fine-tuned VAR outperforms the state-of-the-art DM adaptation strategy DiffFit, that VAR converges quickly, and that DP adaptations of VAR currently suffer from low generation quality and slow convergence.

Significance. If the comparative claims held, this would be a useful empirical data point for the IAR-versus-diffusion debate and would provide the first systematic study of DP fine-tuning for VAR. The released code and the detailed implementation patches for Opacus compatibility are concrete contributions, and the paper is honest about the limitations of its DP compute-cost analysis. However, the headline comparison with DiffFit is undermined by an internal contradiction in Table 1 and by the use of externally quoted baseline FID numbers under a potentially different evaluation protocol; the DP results also contain inconsistencies between tables. The significance is therefore conditional on correcting these issues.

major comments (4)
  1. [Section 3.2 / Table 1] The claim that 'FFT outperforms [DiffFit] across all models and datasets' is contradicted by Table 1 on CUB-200-2011: DiffFit achieves FID 5.48, while VAR-d16 FFT achieves 5.74 and VAR-d20 FFT achieves 5.58, so both VAR FFT variants are worse than DiffFit on that dataset. Additionally, on Oxford Flowers, VAR-d20 LoRA (11.16) beats VAR-d20 FFT (11.65), contradicting the statement in the same section that FFT performs best according to Table 1. The conclusion repeats the unsupported comparative statement. The authors should revise the comparative claims or provide corrected results.
  2. [Section 3.1 / Table 1] The FID comparison with DiT-XL-2 is not apples-to-apples: the DiT-XL-2 and DiffFit numbers are quoted from prior work, whereas the VAR numbers are computed with the protocol described in Section 3.1, which generates as many samples per class as there are test samples and uses a particular FID implementation. Without re-running the diffusion baselines under the same protocol, the favorable comparisons on Food-101, Oxford Flowers, and Stanford Cars are uncertain. The CUB contradiction already shows that the comparison is sensitive to evaluation choices; the authors should either re-run the baselines or explicitly report and discuss the protocol differences.
  3. [Section 4.2 / Tables 2 and 3 / Appendix B] The DP experiments contain an inconsistency: Table 2 reports VAR-d16 LoRA k=128 FID 63.24 and VAR-d20 LoRA k=128 FID 59.29, while Table 3 reports ε=10, k=32 FIDs of 60.24 and 63.38 for the same models; Appendix B lists k=128 as the augmentation multiplicity for DP-ε-10. The ordering between k=32 and k=128 is also non-monotonic across model sizes. Since the augmentation-multiplicity conclusions depend on these comparisons, the authors should state which configuration is the main reported DP result, reconcile the table entries, and explain the non-monotonic behaviour.
  4. [Section 4.2 / Abstract] The paper states that DP adaptations of VAR 'suffer from low generation quality' and 'necessitate further research', but it does not compare against any differentially private diffusion-model baseline (such as DPDM, DP-LDM, or DP-LoRA) under the same evaluation protocol. If the claim is meant only as an absolute statement about VAR's DP performance, that should be stated explicitly; if it is meant as a comparative statement against DM DP adaptations, the missing baseline comparison is a load-bearing gap.
minor comments (6)
  1. [References] The citation 'Gal et al.' in the Introduction is incomplete; it should include the full reference with year and venue for 'An Image is Worth One Word'. The first reference in the bibliography, 'The mnist database of handwritten digits', is also malformed and unattributed.
  2. [Figure 2] The x-axis of Figure 2 has tick labels '1 100 1000 19000 ...' that do not clearly indicate whether a log scale is used and at which step counts the curves are evaluated; please clarify the axis scale and the reporting interval.
  3. [Table 9] In Table 9, the Stanford Cars entry for VAR-d20 FFT reads '59.01' without a percent sign, unlike the other entries; this appears to be a typo.
  4. [Abstract / Section 1] The motivation mentions 'medical data generation' as a downstream task, but no medical dataset is used in the experiments; either add such an experiment or remove this motivation.
  5. [Table 2] Table 2 reports only k=1 for LNTuning, although Appendix B specifies k=128 for DP LNTuning; the missing k=128 LNTuning result should be reported or explicitly explained.
  6. [Section 3.2] The statement that LoRA 'matches FFT's FID performance' is not well supported by Table 1, where, for example, VAR-d20 FFT achieves 5.38 on Food-101 while VAR-d20 LoRA achieves 6.97, and on Stanford Cars the FFT/LoRA gap is 6.31 vs 9.42; please qualify this claim.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the paper is an empirical benchmark whose comparisons use external baselines and measured FID/PFLOps numbers.

full rationale

This paper contains no derivation chain whose conclusion is equivalent to its inputs. The central claims are empirical: FFT and PEFT adaptations of VAR are benchmarked on five downstream datasets, compared against DiT-XL-2 numbers quoted from prior work (Peebles and Xie; Xie et al.), and DP fine-tuning is evaluated on Oxford Flowers. FID values in Table 1, PFLOps in Figure 1, and convergence curves in Figure 2 are all measured outputs, not constructed from the claims they support. The FFT, LoRA, and LNTuning hyperparameters are stated choices, but they are not fitted to a subset of the reported FID scores in a way that forces the comparison outcome, so the 'fitted input called prediction' pattern does not apply. The comparison to DiffFit relies on external baseline numbers, which raises a protocol-comparability concern, but quoting an external baseline is not circularity. The paper cites prior work by non-overlapping authors for methods it implements, and its own contribution is the implementation and measurement; no load-bearing argument reduces to a self-citation. A possible internal inconsistency in Section 3.2 versus Table 1 (DiffFit's FID of 5.48 on CUB-200-2011 being better than the VAR FFT values of 5.74 and 5.58) would be an empirical or reporting error, not a circularity. Therefore the appropriate finding is no significant circularity.

Assumptions & free parameters 7 free parameters · 4 assumptions · 0 invented entities

The paper contributes an empirical benchmark; there are no invented entities and no derivation that hides fitted constants. The free parameters are training hyperparameters that affect the reported FID numbers but are not fitted post hoc to manufacture the headline result.

free parameters (7)
  • LoRA rank r = 16
    Chosen for all LoRA experiments; controls the number of trainable parameters and is known to affect both utility and DP noise sensitivity.
  • LoRA scaling alpha = 2r = 32
    Standard LoRA choice; affects the magnitude of low-rank updates and final FID.
  • Per-dataset learning rates = 1e-4 to 1e-3
    Different learning rates for FFT, LoRA, and LNTuning are listed in Appendix B and materially affect convergence and final FID.
  • Per-dataset epochs = 6 to 50 (non-DP), 80 to 180 (DP)
    Training length was chosen per dataset and per method; DP runs are given many more epochs to compensate for noise.
  • Augmentation multiplicity k = 1, 32, 128
    In DP experiments, k=128 for most runs and k=32 for the epsilon sweep; larger k improves FID modestly at 128x compute.
  • DP gradient clip norm = 0.1 (FFT), 0.5 (LoRA/LNTuning)
    Clipping norm in DP-SGD determines the noise scale and is set differently by method.
  • DP sample rate q = 0.251
    Poisson subsampling rate for Oxford Flowers DP runs; determines per-step batch size and the privacy accounting.
assumptions (4)
  • domain assumption The Opacus library correctly computes per-sample gradients and privacy accounting for DP-SGD under the patched forward function.
    All DP guarantees and FID results depend on the correctness of Opacus accounting and the claim that the forward patch preserves functional parity (Appendix F.4).
  • domain assumption The FID protocol used for VAR-generated images is comparable to the FID numbers quoted for DiT-XL-2 from prior publications.
    The cross-model comparison in Table 1 relies on this equivalence; the paper does not re-run the diffusion baselines.
  • domain assumption The patched buffer and attention implementations are mathematically equivalent to the original VAR implementation.
    The authors state this in Appendix F but provide no direct numerical verification (e.g., identical output tensors on test inputs).
  • domain assumption Pretrained VAR checkpoints and the VQ-GAN tokenizer are fixed and are not updated during adaptation.
    The paper uses ImageNet-1k pretrained VAR models and a pretrained tokenizer; if these were updated, the compute and privacy analysis would differ.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Implementing Adaptations for Vision AutoRegressive Model." pith.science (2026). https://pith.science/paper/JESQV7RJ

@misc{pith2026250711441,
  author       = {Pith},
  title        = {Pith review of: Implementing Adaptations for Vision AutoRegressive Model},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JESQV7RJ}},
  note         = {Machine review of arXiv:2507.11441}
}
read the original abstract

Vision AutoRegressive model (VAR) was recently introduced as an alternative to Diffusion Models (DMs) in image generation domain. In this work we focus on its adaptations, which aim to fine-tune pre-trained models to perform specific downstream tasks, like medical data generation. While for DMs there exist many techniques, adaptations for VAR remain underexplored. Similarly, differentially private (DP) adaptations-ones that aim to preserve privacy of the adaptation data-have been extensively studied for DMs, while VAR lacks such solutions. In our work, we implement and benchmark many strategies for VAR, and compare them to state-of-the-art DM adaptation strategies. We observe that VAR outperforms DMs for non-DP adaptations, however, the performance of DP suffers, which necessitates further research in private adaptations for VAR. Code is available at https://github.com/sprintml/finetuning_var_dp.

Figures

Figures reproduced from arXiv: 2507.11441 by the authors.

Figure 1
Figure 1. Training Compute Cost (PFLOPs) Comparison Across Datasets [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. VARs converge after small amount of training steps. Dataset: CUB-200-2011. Implementation Details The original implementation of the attention operator in VAR requires patches ( [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Extended Training Compute Cost (PFLOPs) Comparison Across Datasets. [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Buffer patching enables skipping register buffer calls to avoid parameter incompatibility with Opacus. F.3. Addressing Gradient Flow Issues in SelfAttention Layers for LoRA & DP-LoRA in VARs The SelfAttention component of the the VAR model contains two base layers we t…
Figure 5
Figure 5. Figure 5: Patching qkv computation to add support for LoRA fine-tuning in VAR models (Tian et al., 2024). To restore compatibility, we override the model’s forward by accepting a single concatenated tensor concat tensor. Before computing the forward pass, we externally concatena…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

36 extracted references · 5 canonical work pages

  1. [1]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION format.date year duplicate empty "emp...

  2. [2]

    http://yann

    The mnist database of handwritten digits. http://yann. lecun. com/exdb/mnist/

    work page

  3. [3]

    Brendan McMahan, Ilya Mironov, Kunal Talwar, and Li Zhang

    Martin Abadi, Andy Chu, Ian Goodfellow, H. Brendan McMahan, Ilya Mironov, Kunal Talwar, and Li Zhang. Deep learning with differential privacy. In Proceedings of the 2016 ACM SIGSAC Conference on Computer and Communications Security, CCS’16. ACM, October 2016. doi:10.1145/2976749.2978318. URL http://dx.doi.org/10.1145/2976749.2978318

    arXiv 2016

  4. [4]

    Intrinsic dimensionality explains the effectiveness of language model fine-tuning, 2020

    Armen Aghajanyan, Luke Zettlemoyer, and Sonal Gupta. Intrinsic dimensionality explains the effectiveness of language model fine-tuning, 2020. URL https://arxiv.org/abs/2012.13255

    arXiv 2020

  5. [5]

    Food-101--mining discriminative components with random forests

    Lukas Bossard, Matthieu Guillaumin, and Luc Van Gool. Food-101--mining discriminative components with random forests. In Computer vision--ECCV 2014: 13th European conference, zurich, Switzerland, September 6-12, 2014, proceedings, part VI 13, pages 446--461. Springer, 2014

    work page 2014

  6. [6]

    Generative pretraining from pixels

    Mark Chen, Alec Radford, Rewon Child, Jeffrey Wu, Heewoo Jun, David Luan, and Ilya Sutskever. Generative pretraining from pixels. In International conference on machine learning, pages 1691--1703. PMLR, 2020

    work page 2020

  7. [7]

    Unlocking high-accuracy differentially private image classification through scale

    Soham De, Leonard Berrada, Jamie Hayes, Samuel L Smith, and Borja Balle. Unlocking high-accuracy differentially private image classification through scale. arXiv preprint arXiv:2204.13650, 2022

    arXiv 2022

  8. [8]

    Imagenet: A large-scale hierarchical image database

    Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. Imagenet: A large-scale hierarchical image database. In 2009 IEEE conference on computer vision and pattern recognition, pages 248--255. Ieee, 2009

    2009

Show all 36 references
  1. [9]

    Differentially private diffusion models, 2023

    Tim Dockhorn, Tianshi Cao, Arash Vahdat, and Karsten Kreis. Differentially private diffusion models, 2023. URL https://arxiv.org/abs/2210.09929

    2023 arXiv

  2. [10]

    Differential privacy

    Cynthia Dwork. Differential privacy. In Michele Bugliesi, Bart Preneel, Vladimiro Sassone, and Ingo Wegener, editors, Automata, Languages and Programming, pages 1--12, Berlin, Heidelberg, 2006. Springer Berlin Heidelberg. ISBN 978-3-540-35908-1

    2006

  3. [11]

    Taming transformers for high-resolution image synthesis, 2020

    Patrick Esser, Robin Rombach, and Björn Ommer. Taming transformers for high-resolution image synthesis, 2020

    2020

  4. [12]

    An image is worth one word: Personalizing text-to-image generation using textual inversion

    Rinon Gal, Yuval Alaluf, Yuval Atzmon, Or Patashnik, Amit Haim Bermano, Gal Chechik, and Daniel Cohen-or. An image is worth one word: Personalizing text-to-image generation using textual inversion. In The Eleventh International Conference on Learning Representations

  5. [13]

    Differentially private diffusion models generate useful synthetic images

    Sahra Ghalebikesabi, Leonard Berrada, Sven Gowal, Ira Ktena, Robert Stanforth, Jamie Hayes, Soham De, Samuel L Smith, Olivia Wiles, and Borja Balle. Differentially private diffusion models generate useful synthetic images. arXiv preprint arXiv:2302.13861, 2023

    2023 arXiv

  6. [14]

    Infinity: Scaling bitwise autoregressive modeling for high-resolution image synthesis, 2024

    Jian Han, Jinlai Liu, Yi Jiang, Bin Yan, Yuqi Zhang, Zehuan Yuan, Bingyue Peng, and Xiaobing Liu. Infinity: Scaling bitwise autoregressive modeling for high-resolution image synthesis, 2024. URL https://arxiv.org/abs/2412.04431

    2024 arXiv

  7. [15]

    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 Conference on Neural Information Processing Systems (NeurIPS), page 6629–6640, 2017

    2017

  8. [16]

    Lora: Low-rank adaptation of large language models

    Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, Weizhu Chen, et al. Lora: Low-rank adaptation of large language models. ICLR, 1 0 (2): 0 3, 2022

    2022

  9. [17]

    Diffusion models in medical imaging: A comprehensive survey

    Amirhossein Kazerouni, Ehsan Khodapanah Aghdam, Moein Heidari, Reza Azad, Mohsen Fayyaz, Ilker Hacihaliloglu, and Dorit Merhof. Diffusion models in medical imaging: A comprehensive survey. Medical Image Analysis, 88: 0 102846, 2023. ISSN 1361-8415. doi:https://doi.org/10.1016/...

    2023

  10. [18]

    3d object representations for fine-grained categorization

    Jonathan Krause, Michael Stark, Jia Deng, and Li Fei-Fei. 3d object representations for fine-grained categorization. In 2013 IEEE International Conference on Computer Vision Workshops, pages 554--561, 2013. doi:10.1109/ICCVW.2013.77

    2013 doi

  11. [19]

    Improved precision and recall metric for assessing generative models

    Tuomas Kynk\" a \" a nniemi, Tero Karras, Samuli Laine, Jaakko Lehtinen, and Timo Aila. Improved precision and recall metric for assessing generative models. In H. Wallach, H. Larochelle, A. Beygelzimer, F. d Alch\' e -Buc, E. Fox, and R. Garnett, editors, Advances in Neural I...

    2019

  12. [20]

    Liu, Saiyue Lyu, Margarita Vinaroz, and Mijung Park

    Michael F. Liu, Saiyue Lyu, Margarita Vinaroz, and Mijung Park. Differentially private latent diffusion models, 2024. URL https://arxiv.org/abs/2305.15759

    2024 arXiv

  13. [21]

    Automated flower classification over a large number of classes

    Maria-Elena Nilsback and Andrew Zisserman. Automated flower classification over a large number of classes. In Indian Conference on Computer Vision, Graphics and Image Processing, Dec 2008

    2008

  14. [22]

    Cats and dogs

    Omkar M Parkhi, Andrea Vedaldi, Andrew Zisserman, and CV Jawahar. Cats and dogs. In 2012 IEEE conference on computer vision and pattern recognition, pages 3498--3505. IEEE, 2012

    2012

  15. [23]

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

    A Paszke. Pytorch: An imperative style, high-performance deep learning library. arXiv preprint arXiv:1912.01703, 2019

    1912 arXiv

  16. [24]

    Scalable diffusion models with transformers

    William Peebles and Saining Xie. Scalable diffusion models with transformers. In Proceedings of the IEEE/CVF international conference on computer vision, pages 4195--4205, 2023

    2023

  17. [25]

    Language models are unsupervised multitask learners

    Alec Radford, Jeff Wu, Rewon Child, David Luan, Dario Amodei, and Ilya Sutskever. Language models are unsupervised multitask learners. 2019

    2019

  18. [26]

    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 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 10674--10685, 2022. doi:10.1109/CVPR52688.2022.01042

    2022

  19. [27]

    Dreambooth: Fine tuning text-to-image diffusion models for subject-driven generation

    Nataniel Ruiz, Yuanzhen Li, Varun Jampani, Yael Pritch, Michael Rubinstein, and Kfir 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...

    2023

  20. [28]

    Mehdi S. M. Sajjadi, Olivier Bachem, Mario Lucic, Olivier Bousquet, and Sylvain Gelly. Assessing generative models via precision and recall, 2018. URL https://arxiv.org/abs/1806.00035

    2018 arXiv

  21. [29]

    Visual autoregressive modeling: Scalable image generation via next-scale prediction, 2024

    Keyu Tian, Yi Jiang, Zehuan Yuan, Bingyue Peng, and Liwei Wang. Visual autoregressive modeling: Scalable image generation via next-scale prediction, 2024. URL https://arxiv.org/abs/2404.02905

    2024 arXiv

  22. [30]

    Differentially private fine-tuning of diffusion models, 2024

    Yu-Lin Tsai, Yizhe Li, Zekai Chen, Po-Yu Chen, Chia-Mu Yu, Xuebin Ren, and Francois Buet-Golfouse. Differentially private fine-tuning of diffusion models, 2024. URL https://arxiv.org/abs/2406.01355

    2024 arXiv

  23. [31]

    Attention is all you need

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, ukasz Kaiser, and Illia Polosukhin. Attention is all you need. Advances in neural information processing systems, 30, 2017

    2017

  24. [32]

    The caltech-ucsd birds-200-2011 dataset

    Catherine Wah, Steve Branson, Peter Welinder, Pietro Perona, and Serge Belongie. The caltech-ucsd birds-200-2011 dataset. 2011

    2011

  25. [33]

    Difffit: Unlocking transferability of large diffusion models via simple parameter-efficient fine-tuning, 2023

    Enze Xie, Lewei Yao, Han Shi, Zhili Liu, Daquan Zhou, Zhaoqiang Liu, Jiawei Li, and Zhenguo Li. Difffit: Unlocking transferability of large diffusion models via simple parameter-efficient fine-tuning, 2023. URL https://arxiv.org/abs/2304.06648

    2023 arXiv

  26. [34]

    Parameter-efficient fine-tuning methods for pretrained language models: A critical review and assessment, 2023

    Lingling Xu, Haoran Xie, Si-Zhao Joe Qin, Xiaohui Tao, and Fu Lee Wang. Parameter-efficient fine-tuning methods for pretrained language models: A critical review and assessment, 2023. URL https://arxiv.org/abs/2312.12148

    2023 arXiv

  27. [35]

    Randomized autoregressive visual generation

    Qihang Yu, Ju He, Xueqing Deng, Xiaohui Shen, and Liang-Chieh Chen. Randomized autoregressive visual generation. arXiv preprint arXiv:2411.00776, 2024

    2024 arXiv

  28. [36]

    Tuning layernorm in attention: Towards efficient multi-modal llm finetuning, 2023

    Bingchen Zhao, Haoqin Tu, Chen Wei, Jieru Mei, and Cihang Xie. Tuning layernorm in attention: Towards efficient multi-modal llm finetuning, 2023. URL https://arxiv.org/abs/2312.11420

    2023 arXiv

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.