Pith. sign in

REVIEW 3 major objections 6 minor 66 references

LoFT: LoRA-fused Training Dataset Generation with Few-shot Guidance

T0 review · 3 major / 6 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read The paper claims that LoFT-generated synthetic data yields higher classification accuracy than prior synthetic-dataset methods, with the gap widening as the dataset grows.

desk verdict LoFT is a well-executed few-shot synthetic data generation method with a genuinely new per-image LoRA fusion mechanism; the headline claim needs scoping and the fusion weight lambda is chosen on ImageNet, but the core contribution is solid. read the letter →

arxiv 2505.11703 v1 pith:LPCWHLJG submitted 2025-05-16 cs.CV

Jae Myung Kim , Stephan Alaniz , Cordelia Schmid , Zeynep Akata This is my paper · ORCID
classification cs.CV
keywords syntheticdatagenerationLoRAfew-shotlearningdiffusionmodelsCLIPfine-tuningdatasetfidelityanddiversityimageclassification
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper introduces LoFT, a way to generate synthetic training images from a few real examples per class. Instead of fine-tuning a diffusion model on all images of a class together, LoFT trains a separate LoRA adapter on each individual real image, then generates new images by interpolating the adapters of two same-class images at equal weight. The paper's central empirical claim is that classifiers fine-tuned on LoFT-generated data outperform classifiers trained on data from previous zero-shot and few-shot generation methods, and that the advantage grows as the synthetic dataset is scaled up. If correct, this makes few-shot guided synthetic data a more reliable route to improving image classifiers without collecting large labeled datasets.

What carries the argument

The load-bearing object is a set of per-image LoRA adapters: for each real image $x_i$, the diffusion model's attention linear layers are updated by a low-rank matrix $\Delta W^{(i)} = B^{(i)} A^{(i)}$ with rank 2, trained to reproduce that single image. Generation then replaces the adapter with an interpolation $h_{\text{out}} = W h_{\text{in}} + \lambda\Delta W^{(i)} h_{\text{in}} + (1-\lambda)\Delta W^{(j)} h_{\text{in}}$ for two same-class images and $\lambda = 0.5$. The per-image overfitting preserves fine-grained details, giving fidelity, while interpolating between two adapters varies appearance and composition, giving diversity; random seeds add further variation.

What would settle it

Run the paper's CLIP fine-tuning protocol on a held-out fine-grained benchmark not used in the paper, with 16-shot guidance and 500 generated images per class; the central claim fails if a LoFT-trained model does not beat CaptionPrompt and DataDream. A direct mechanism check is to measure whether $\lambda = 0.5$ fused generations actually interpolate: if per-class FID or feature statistics of fused images are essentially identical to one of the two source images across many sampled pairs, the diversity mechanism is not working as claimed.

Watch

Extended reading notes

Core claim

LoFT's central claim is that per-image LoRA fine-tuning followed by weight-space fusion resolves the fidelity-diversity trade-off that limits earlier synthetic data methods. On ImageNet, LoFT-generated data beats ClassPrompt, CaptionPrompt, and DataDream at every scale from 50K to 1M images and at every guidance level from 8- to 64-shot, with accuracy rising from 71.02% to 72.24% in the 16-shot setting as the dataset grows. On nine fine-grained benchmarks it reaches an average of 84.7% versus 83.2% for DataDream with 500 images per class, and it outperforms few-shot learning baselines when synthetic and real data are combined, reaching 87.2% average accuracy versus 82.1% for the next best baseline. The analysis attributes the gain to high recognizability, that is fidelity, together with sufficient diversity, where fusing two LoRA adapters at $\lambda = 0.5$ yields images that inherit attributes from both source images while remaining recognizable as the class.

Load-bearing premise

The method assumes that linearly averaging the LoRA weights of two real images produces new images that combine both images' features while staying inside the class distribution; if the interpolation instead creates artifacts, near-duplicates of one source, or out-of-distribution hybrids, the claimed fidelity-diversity advantage collapses.

Editorial extensions

If this is right

  • A practitioner with only 8 to 64 labeled images per class can generate a synthetic training set that improves CLIP fine-tuning more than the real few-shot images alone.
  • Unlike template-prompt generation, LoFT's advantage grows as more synthetic images are added, at least up to 1M images on ImageNet.
  • On fine-grained benchmarks, 500 LoFT images per class plus the few-shot real data outperforms strong few-shot CLIP methods, suggesting synthetic data can substitute for large amounts of manual annotation.
  • The best configuration is equal-weight fusion of exactly two adapters, so the specific choice of pairwise interpolation with $\lambda = 0.5$ is central to the method's success.
  • At higher shot counts, LoFT data also improves training a ResNet50 from scratch on ImageNet, extending the finding beyond fine-tuning a pre-trained CLIP model.

Reading between the lines

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

  • Beyond the paper's classification benchmarks, per-image LoRA fusion could transfer to detection, segmentation, or long-tail recognition, where per-instance detail and in-distribution generation matter most.
  • The paper observes an inverse correlation between recognizability and diversity, which suggests a hybrid that mixes high-fidelity LoFT images with more diverse caption-prompt images might push accuracy further than either method alone.
  • Because LoFT improves as the number of guidance images grows, it may be especially useful for rare classes with only a handful of examples, where individual-image features are decisive.
  • A natural testable extension is weighted barycenters of three or more adapters or sampling $\lambda$ around 0.5; the paper's ablations suggest two-adapter equal-weight fusion is hard to beat, but that conclusion is currently empirical rather than principled.
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

3 major / 6 minor

Summary. The paper introduces LoFT, a few-shot guided synthetic dataset generation method for image classification. LoFT fine-tunes a separate low-rank adapter (LoRA) for each real image in a small support set, freezing the base Stable Diffusion model, and then generates synthetic images by randomly selecting two adapters of the same class and linearly interpolating their weight matrices with a fixed fusion weight λ=0.5. The authors claim this preserves instance-level fidelity while improving diversity relative to per-image reproduction. The method is evaluated by fine-tuning CLIP ViT-B/16 on synthetic ImageNet subsets ranging from 0.05M to 1M images under 8/16/32/64-shot guidance, on nine fine-grained datasets with 16-shot guidance and 500 images per class, and by training ResNet50 from scratch on ImageNet. The paper also compares against ClassPrompt, CaptionPrompt, and DataDream, analyzes fidelity and diversity, and ablates the fusion representation and the fusion weight λ. The main result is that LoFT achieves the highest average accuracy in the CLIP fine-tuning protocol and generally scales better than the baselines.

Significance. If the central claim holds, LoFT is a practically useful method for few-shot guided synthetic dataset generation, with a simple and modular design. The paper's strengths include the breadth of the evaluation (10 datasets, 4 k-shot settings, dataset sizes up to 1M, and an additional from-scratch protocol), the release of code, and the explicit per-class fidelity/diversity analysis. The per-image LoRA fusion idea is novel and clearly explained, and the qualitative results support the plausibility of the key assumption that weight-space interpolation yields in-distribution feature mixes. However, the strong 'consistently outperforms' claim is not fully supported: the main hyperparameter λ is selected on the headline ImageNet benchmark and has a large effect, one fusion ablation uses a different generative backbone, and some benchmark results contradict the unqualified wording. These issues are addressable in revision.

major comments (3)
  1. [§4.5.2, Table 4] The fusion weight λ is fixed to 0.5 for all the main experiments in Figure 3 and Tables 1-2, but this value is selected via an ablation on ImageNet, the same dataset used for the headline scaling curves. Table 4 shows the effect is large: at 0.5M synthetic images with from-scratch ResNet50, λ=0.5 gives 45.41% versus 39.18% for λ=0.7 and 30.85% for λ=1.0. These gaps are much larger than the differences between methods in Figure 3 and Table 1. To support the universal claim, the paper should either demonstrate on held-out datasets (e.g., a DTD or Aircraft λ sweep) that λ=0.5 is uniformly optimal, or treat λ as a validation-selected hyperparameter, or provide a formal or empirical justification for the symmetric choice. As written, the possibility that the headline superiority is an artifact of target-set hyperparameter tuning is not excluded.
  2. [§4.5.1, Table 3] The image-embedding fusion baseline is evaluated with Stable Diffusion 2.1-unclip, a different generative backbone from the Stable Diffusion 2.1 used by LoFT. This confounds the fusion representation with the generative model architecture, so the comparison does not isolate the effect of fusing LoRA weights versus fusing image embeddings. Please either align the backbones (e.g., use unclip for all variants or use an image-conditioned variant of SD2.1 for the baseline) or explicitly acknowledge the confound and soften the conclusion drawn from this table.
  3. [Abstract and §4.2.1] The claim that training on LoFT-generated data 'consistently outperforms other synthetic dataset methods' is stronger than the reported results. In Table 1, DataDream achieves higher accuracy on Stanford Cars (90.5 vs 89.3) and Food101 (86.5 vs 86.0), and in Appendix B CaptionPrompt outperforms LoFT at 8-shot for from-scratch training. The paper should rephrase the headline claim to reflect that LoFT achieves the best average accuracy and wins on most but not all benchmarks and settings.
minor comments (6)
  1. [Figure 3 and Figure 6] The scaling curves are shown without error bars or repeated-seed variance; given the small gaps between methods near the top of the y-axis (e.g., less than 1 point between LoFT and CaptionPrompt in the 8-shot panel), a single run makes it hard to assess whether the differences are significant. Please add multiple seeds and error bars or a statistical test.
  2. [§3.2, Eq. (4)] The notation in Eq. (4) is slightly ambiguous: 'λ∆W(i)hin + (1−λ)∆W(j)hin' should be parenthesized as λ(∆W(i)hin) + (1−λ)(∆W(j)hin) or defined as the interpolation of weight matrices before application. This is easily clarified.
  3. [§4.2.1, Table 1] The 'Avg' column is computed over nine datasets, but the per-dataset standard deviations are reported only for the last digit; consider adding the standard error of the mean or clarifying the number of seeds.
  4. [§4.5.2, Table 4] The row labels such as '[0.5,0.25,0.25]' are not defined in the caption; please state that these are the weights for fusing three LoRA adapters and how they are normalized.
  5. [§3.1] The sentence 'we assume to have access' should be 'we assume access'; also, 'the diffusion model learns to overfit to a single image, generating it exclusively' is a bit imprecise because the generation still depends on the prompt and random seed. Please rephrase.
  6. [Limitations] The limitation paragraph is candid about storage and compute, but it does not mention the sensitivity to λ or the target-set selection issue; a sentence acknowledging this would help the reader calibrate the universality of the results.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity; LoFT's central claim is empirically validated against external benchmarks and not constructed from its inputs.

full rationale

The derivation chain is empirical and self-contained. LoFT's pipeline is defined independently: per-image LoRA fine-tuning (Eq. 3) and weight interpolation (Eq. 4), with no parameter fitted to the quantities later claimed as predictions. The reported comparisons are held-out validation accuracies from externally pretrained CLIP ViT-B/16 and from-scratch ResNet50; recognizability and diversity metrics use a pretrained ImageNet ViT classifier, so the fidelity claims are not defined in terms of LoFT's own outputs. DataDream, the closest baseline, is prior work by overlapping authors, but it is used only as a comparative baseline with stated hyperparameters, not as a premise or uniqueness theorem, so it does not carry the argument. The lambda = 0.5 choice is selected on ImageNet in Table 4 and then reused in the main ImageNet experiments, which is a generalization/test-set-selection concern rather than circularity: it does not make LoFT's accuracy equal to the selection criterion by construction. Therefore no load-bearing step reduces to its inputs.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The central claim depends on a handful of hyperparameters (lambda, rank, guidance scale, prompt template) and on the untested assumption that LoRA weight interpolation behaves semantically. No new theoretical entities are introduced.

free parameters (5)
  • fusion weight lambda = 0.5
    Selected via ImageNet accuracy in Table 4; controls how much each source image contributes to the generated image.
  • LoRA rank for per-image adapters = 2
    Chosen as sufficient for single-image adaptation; no hyperparameter search reported.
  • guidance scale = 2.0
    Used for all generation methods, following DataDream; affects fidelity/diversity trade-off.
  • CLIP fine-tuning LoRA rank = 16
    Standard rank for fine-tuning CLIP in prior work; fixed across methods.
  • prompt template = 'a photo of a {l}'
    Template prompt used for all methods; class name only, no additional attributes.
assumptions (4)
  • domain assumption Stable Diffusion 2.1 can generate recognizable images for the 1,000 ImageNet classes and the fine-grained classes used in evaluation.
    The method inherits the generator's capability; if the base model cannot render a class, LoFT cannot fix it.
  • ad hoc to paper A LoRA adapter with rank 2 can overfit a diffusion model to a single image sufficiently to reproduce it.
    Core to Phase 1 of LoFT; no theoretical guarantee, empirical evidence via fidelity metrics.
  • ad hoc to paper Linear interpolation of two LoRA weight matrices yields a semantically valid feature mix.
    Core to Phase 2 of LoFT (Eq. 4); motivated by prior LoRA morphing work but not derived.
  • domain assumption Pre-trained CLIP ViT-B/16 is a suitable base for fine-tuning with synthetic data.
    The paper's main evaluation protocol; assumed to transfer to other backbones.

how reviews work

0 comments
Cite this review

Pith. "Pith review of LoFT: LoRA-fused Training Dataset Generation with Few-shot Guidance." pith.science (2026). https://pith.science/paper/LPCWHLJG

@misc{pith2026250511703,
  author       = {Pith},
  title        = {Pith review of: LoFT: LoRA-fused Training Dataset Generation with Few-shot Guidance},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/LPCWHLJG}},
  note         = {Machine review of arXiv:2505.11703}
}
read the original abstract

Despite recent advances in text-to-image generation, using synthetically generated data seldom brings a significant boost in performance for supervised learning. Oftentimes, synthetic datasets do not faithfully recreate the data distribution of real data, i.e., they lack the fidelity or diversity needed for effective downstream model training. While previous work has employed few-shot guidance to address this issue, existing methods still fail to capture and generate features unique to specific real images. In this paper, we introduce a novel dataset generation framework named LoFT, LoRA-Fused Training-data Generation with Few-shot Guidance. Our method fine-tunes LoRA weights on individual real images and fuses them at inference time, producing synthetic images that combine the features of real images for improved diversity and fidelity of generated data. We evaluate the synthetic data produced by LoFT on 10 datasets, using 8 to 64 real images per class as guidance and scaling up to 1000 images per class. Our experiments show that training on LoFT-generated data consistently outperforms other synthetic dataset methods, significantly increasing accuracy as the dataset size increases. Additionally, our analysis demonstrates that LoFT generates datasets with high fidelity and sufficient diversity, which contribute to the performance improvement. The code is available at https://github.com/ExplainableML/LoFT.

Figures

Figures reproduced from arXiv: 2505.11703 by the authors.

Figure 2
Figure 2. LoFT pipeline. In the first phase, given a few real images per class, we adapt a diffusion model to each image using LoRA. In [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Classification accuracy on ImageNet when fine-tuning CLIP on synthetic data generated from different methods at different [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Per-class analysis on synthetic datasets generated from different methods. The color indicates a log-likelihood of the ImageNet [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figures from the paper (10 more)
Figure 5
Figure 5. Figure 5: Qualitative examples for the classes Acoustic guitar and [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: Classification accuracy on ImageNet when training ResNet50 from scratch on synthetic data generated from different methods [PITH_FULL_IMAGE:figures/full_fig_p013_6.png]
Figure 7
Figure 7. Figure 7: Scaling the number of synthetic data on Aircraft and DTD [PITH_FULL_IMAGE:figures/full_fig_p013_7.png]
Figure 9
Figure 9. Figure 9: Ablation study of qualitative results on [PITH_FULL_IMAGE:figures/full_fig_p014_9.png]
Figure 10
Figure 10. Figure 10: Per-class analysis of recognizability and alignment in synthetic datasets generated from different methods. The color indicates [PITH_FULL_IMAGE:figures/full_fig_p015_10.png]
Figure 11
Figure 11. Figure 11: Per-class analysis of diversity and alignment in synthetic datasets generated from different methods. The color indicates a [PITH_FULL_IMAGE:figures/full_fig_p015_11.png]
Figure 12
Figure 12. Figure 12: Qualitative examples for the classes Hourglass, Hard disk drive, Joystick, and Weighing scale. [PITH_FULL_IMAGE:figures/full_fig_p016_12.png]
Figure 13
Figure 13. Figure 13: Qualitative examples for the classes Carved Pumpkin, Diaper, Swing, and iPod. [PITH_FULL_IMAGE:figures/full_fig_p017_13.png]
Figure 14
Figure 14. Figure 14: Ablation study of qualitative results on [PITH_FULL_IMAGE:figures/full_fig_p018_14.png]
Figure 15
Figure 15. Figure 15: Qualitative results of our LoFT method on Aircraft and Cars datasets. [PITH_FULL_IMAGE:figures/full_fig_p019_15.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

66 extracted references · 37 canonical work pages

  1. [1]

    Self-consuming gen- erative models go mad

    Sina Alemohammad, Josue Casco-Rodriguez, Lorenzo Luzi, Ahmed Imtiaz Humayun, Hossein Babaei, Daniel LeJeune, Ali Siahkoohi, and Richard Baraniuk. Self-consuming gen- erative models go mad. In ICLR, 2023. 2

    work page 2023

  2. [2]

    Shekoofeh Azizi, Simon Kornblith, Chitwan Saharia, Mo- hammad Norouzi, and David J. Fleet. Synthetic data from diffusion models improves imagenet classification. TMLR,

    work page

  3. [3]

    Leaving reality to imag- ination: Robust classification via generated datasets

    Hritik Bansal and Aditya Grover. Leaving reality to imag- ination: Robust classification via generated datasets. arXiv preprint arXiv:2302.02503, 2023. 2

    arXiv 2023

  4. [4]

    On the stability of iterative retraining of generative models on their own data

    Quentin Bertrand, Avishek Joey Bose, Alexandre Duplessis, Marco Jiralerspong, and Gauthier Gidel. On the stability of iterative retraining of generative models on their own data. In ICLR, 2024. 2

    work page 2024

  5. [5]

    Improving image generation with better captions

    James Betker, Gabriel Goh, Li Jing, Tim Brooks, Jianfeng Wang, Linjie Li, Long Ouyang, Juntang Zhuang, Joyce Lee, Yufei Guo, et al. Improving image generation with better captions. Computer Science. https://cdn. openai. com/papers/dall-e-3. pdf, 2(3):8, 2023. 1, 2

    2023

  6. [6]

    Paligemma: A versatile 3b vlm for trans- fer

    Lucas Beyer, Andreas Steiner, Andr ´e Susano Pinto, Alexan- der Kolesnikov, Xiao Wang, Daniel Salz, Maxim Neumann, Ibrahim Alabdulmohsin, Michael Tschannen, Emanuele Bugliarello, et al. Paligemma: A versatile 3b vlm for trans- fer. arXiv preprint arXiv:2407.07726, 2024. 4

    arXiv 2024

  7. [7]

    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 ECCV, pages 446–461. Springer, 2014. 6

    2014

  8. [8]

    An empiri- cal study of training self-supervised vision transformers

    Xinlei Chen, Saining Xie, and Kaiming He. An empiri- cal study of training self-supervised vision transformers. In ICCV, 2021. 6

    work page 2021

Show all 66 references
  1. [9]

    Cimpoi, S

    M. Cimpoi, S. Maji, I. Kokkinos, S. Mohamed, , and A. Vedaldi. Describing textures in the wild. In CVPR, 2014. 6

    2014

  2. [10]

    Turrisi da Costa, Nicola Dall’Asen, Yiming Wang, Nicu Sebe, and Elisa Ricci

    Victor G. Turrisi da Costa, Nicola Dall’Asen, Yiming Wang, Nicu Sebe, and Elisa Ricci. Diversified in-domain synthesis with efficient fine-tuning for few-shot classification, 2023. 1, 2, 5

    2023

  3. [11]

    Interpreting the weight space of customized dif- fusion models

    Amil Dravid, Yossi Gandelsman, Kuan-Chieh Wang, Rameen Abdal, Gordon Wetzstein, Alexei A Efros, and Kfir Aberman. Interpreting the weight space of customized dif- fusion models. arXiv preprint arXiv:2406.09413, 2024. 2

    2024 arXiv

  4. [12]

    Dream the impossible: Outlier imagination with diffusion models

    Xuefeng Du, Yiyou Sun, Jerry Zhu, and Yixuan Li. Dream the impossible: Outlier imagination with diffusion models. NeurIPS, 2023. 2

    2023

  5. [13]

    Gonzalez, and Trevor Darrell

    Lisa Dunlap, Alyssa Umino, Han Zhang, Jiezhi Yang, Joseph E. Gonzalez, and Trevor Darrell. Diversify your vi- sion datasets with automatic diffusion-based augmentation,

  6. [14]

    Scaling laws of synthetic images for model training

    Lijie Fan, Kaifeng Chen, Dilip Krishnan, Dina Katabi, Phillip Isola, and Yonglong Tian. Scaling laws of synthetic images for model training... for now. In CVPR, 2024. 1, 2, 3, 4, 6

    2024

  7. [15]

    Dreamda: Generative data augmentation with diffusion mod- els

    Yunxiang Fu, Chaoqi Chen, Yu Qiao, and Yizhou Yu. Dreamda: Generative data augmentation with diffusion mod- els. arXiv preprint arXiv:2403.12803, 2024. 2

    2024 arXiv

  8. [16]

    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 ICLR, 2022. 2, 8

    2022

  9. [17]

    Synthclip: Are we ready for a fully synthetic clip training? arXiv preprint arXiv:2402.01832, 2024

    Hasan Abed Al Kader Hammoud, Hani Itani, Fabio Pizzati, Philip Torr, Adel Bibi, and Bernard Ghanem. Synthclip: Are we ready for a fully synthetic clip training? arXiv preprint arXiv:2402.01832, 2024. 2 9

    2024 arXiv

  10. [18]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In CVPR,

  11. [19]

    Is syn- thetic data from generative models ready for image recogni- tion? In ICLR, 2023

    Ruifei He, Shuyang Sun, Xin Yu, Chuhui Xue, Wenqing Zhang, Philip Torr, Song Bai, and XIAOJUAN QI. Is syn- thetic data from generative models ready for image recogni- tion? In ICLR, 2023. 1, 2, 3

    2023

  12. [20]

    Eurosat: A novel dataset and deep learn- ing benchmark for land use and land cover classification

    Patrick Helber, Benjamin Bischke, Andreas Dengel, and Damian Borth. Eurosat: A novel dataset and deep learn- ing benchmark for land use and land cover classification. AEROS, 2019. 6

    2019

  13. [21]

    Feedback-guided data synthesis for imbalanced classifica- tion

    Reyhane Askari Hemmat, Mohammad Pezeshki, Florian Bordes, Michal Drozdzal, and Adriana Romero-Soriano. Feedback-guided data synthesis for imbalanced classifica- tion. arXiv preprint arXiv:2310.00158, 2023. 2

    2023 arXiv

  14. [22]

    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. NeurIPS, 2017. 12

    2017

  15. [23]

    Denoising diffu- sion probabilistic models

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising diffu- sion probabilistic models. NeurIPS, 2020. 1

    2020

  16. [24]

    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. 2, 4

    2022

  17. [25]

    Datadream: Few-shot guided dataset generation

    Jae Myung Kim, Jessica Bader, Stephan Alaniz, Cordelia Schmid, and Zeynep Akata. Datadream: Few-shot guided dataset generation. ECCV, 2024. 1, 2, 3, 4, 5

    2024

  18. [26]

    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 ICCV workshop, pages 554–561, 2013. 6

    2013

  19. [27]

    Image captions are natural prompts for text-to-image models

    Shiye Lei, Hao Chen, Sen Zhang, Bo Zhao, and Dacheng Tao. Image captions are natural prompts for text-to-image models. arXiv preprint arXiv:2307.08526, 2023. 2

    2023 arXiv

  20. [28]

    Caltech 101, 2022

    Fei-Fei Li, Marco Andreeto, Marc’Aurelio Ranzato, and Pietro Perona. Caltech 101, 2022. 6

    2022

  21. [29]

    Ex- plore the power of synthetic data on few-shot object detec- tion

    Shaobo Lin, Kun Wang, Xingyu Zeng, and Rui Zhao. Ex- plore the power of synthetic data on few-shot object detec- tion. In CVPR, 2023. 2

    2023

  22. [30]

    Does feasi- bility matter? understanding the impact of feasibility on syn- thetic training data

    Yiwen Liu, Jessica Bader, and Jae Myung Kim. Does feasi- bility matter? understanding the impact of feasibility on syn- thetic training data. In CVPR Workshop in Synthetic Data for Computer Vision, 2025. 2

    2025

  23. [31]

    Decoupled weight decay regularization

    Ilya Loshchilov and Frank Hutter. Decoupled weight decay regularization. In ICLR, 2018. 5

    2018

  24. [32]

    S. Maji, J. Kannala, E. Rahtu, M. Blaschko, and A. Vedaldi. Fine-grained visual classification of aircraft. Technical re- port, 2013. 6

    2013

  25. [33]

    Glide: Towards photorealistic image generation and editing with text-guided diffusion models.arXiv preprint arXiv:2112.10741, 2021

    Alex Nichol, Prafulla Dhariwal, Aditya Ramesh, Pranav Shyam, Pamela Mishkin, Bob McGrew, Ilya Sutskever, and Mark Chen. Glide: Towards photorealistic image generation and editing with text-guided diffusion models.arXiv preprint arXiv:2112.10741, 2021. 2

    2021 arXiv

  26. [34]

    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 In- dian Conference on Computer Vision, Graphics and Image Processing, 2008. 6

    2008

  27. [35]

    Parkhi, Andrea Vedaldi, Andrew Zisserman, and C

    Omkar M. Parkhi, Andrea Vedaldi, Andrew Zisserman, and C. V . Jawahar. Cats and dogs. InCVPR, 2012. 6

    2012

  28. [36]

    Sdxl: Improving latent diffusion models 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 models for high-resolution image synthesis. In ICLR, 2023. 1, 2

    2023

  29. [37]

    Controlling text-to-image diffusion by orthogo- nal finetuning

    Zeju Qiu, Weiyang Liu, Haiwen Feng, Yuxuan Xue, Yao Feng, Zhen Liu, Dan Zhang, Adrian Weller, and Bernhard Sch¨olkopf. Controlling text-to-image diffusion by orthogo- nal finetuning. NeurIPS, 2023. 2

    2023

  30. [38]

    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. 5, 6

    2021

  31. [39]

    Hierarchical text-conditional image gener- ation with clip latents

    Aditya Ramesh, Prafulla Dhariwal, Alex Nichol, Casey Chu, and Mark Chen. Hierarchical text-conditional image gener- ation with clip latents. arXiv preprint arXiv:2204.06125, 1 (2):3, 2022. 1, 2

    2022 arXiv

  32. [40]

    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. 1, 2, 3, 4

    2022

  33. [41]

    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 CVPR, pages 22500–22510, 2023. 2

    2023

  34. [42]

    Hyperdreambooth: Hypernetworks for fast personalization of text-to-image models

    Nataniel Ruiz, Yuanzhen Li, Varun Jampani, Wei Wei, Tingbo Hou, Yael Pritch, Neal Wadhwa, Michael Rubinstein, and Kfir Aberman. Hyperdreambooth: Hypernetworks for fast personalization of text-to-image models. InCVPR, 2024. 2

    2024

  35. [43]

    Imagenet large scale visual recognition challenge

    Olga Russakovsky, Jia Deng, Hao Su, Jonathan Krause, San- jeev Satheesh, Sean Ma, Zhiheng Huang, Andrej Karpathy, Aditya Khosla, Michael Bernstein, et al. Imagenet large scale visual recognition challenge. IJCV, 2015. 5

    2015

  36. [44]

    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. NeurIPS, 2022. 1, 2

    2022

  37. [45]

    Fake it till you make it: Learning trans- ferable representations from synthetic imagenet clones

    Mert Bulent Sariyildiz, Karteek Alahari, Diane Larlus, and Yannis Kalantidis. Fake it till you make it: Learning trans- ferable representations from synthetic imagenet clones. In CVPR, 2023. 1, 2, 3

    2023

  38. [46]

    Synth2: Boosting visual-language models with synthetic captions and image embeddings

    Sahand Sharifzadeh, Christos Kaplanis, Shreya Pathak, Dharshan Kumaran, Anastasija Ilic, Jovana Mitrovic, Charles Blundell, and Andrea Banino. Synth2: Boosting visual-language models with synthetic captions and image embeddings. arXiv preprint arXiv:2403.07750, 2024. 2

    2024 arXiv

  39. [47]

    Instant- booth: Personalized text-to-image generation without test- time finetuning

    Jing Shi, Wei Xiong, Zhe Lin, and Hyun Joon Jung. Instant- booth: Personalized text-to-image generation without test- time finetuning. In CVPR, 2024. 2

    2024

  40. [48]

    Fill-up: Balancing long-tailed data with generative models

    Joonghyuk Shin, Minguk Kang, and Jaesik Park. Fill-up: Balancing long-tailed data with generative models. arXiv preprint arXiv:2306.07200, 2023. 2

    2023 arXiv

  41. [49]

    Diversity is definitely needed: 10 Improving model-agnostic zero-shot classification via stable diffusion, 2023

    Jordan Shipard, Arnold Wiliem, Kien Nguyen Thanh, Wei Xiang, and Clinton Fookes. Diversity is definitely needed: 10 Improving model-agnostic zero-shot classification via stable diffusion, 2023. 1, 2

    2023

  42. [50]

    Semantic-aware data augmentation for text-to-image synthesis

    Zhaorui Tan, Xi Yang, and Kaizhu Huang. Semantic-aware data augmentation for text-to-image synthesis. In AAAI,

  43. [51]

    Amu-tuning: Effective logit bias for clip-based few-shot learning

    Yuwei Tang, Zhenyi Lin, Qilong Wang, Pengfei Zhu, and Qinghua Hu. Amu-tuning: Effective logit bias for clip-based few-shot learning. In CVPR, 2024. 6

    2024

  44. [52]

    Stablerep: Synthetic images from text-to- image models make strong visual representation learners

    Yonglong Tian, Lijie Fan, Phillip Isola, Huiwen Chang, and Dilip Krishnan. Stablerep: Synthetic images from text-to- image models make strong visual representation learners. NeurIPS, 2023. 2

    2023

  45. [53]

    Learning vision from mod- els rivals learning vision from data

    Yonglong Tian, Lijie Fan, Kaifeng Chen, Dina Katabi, Dilip Krishnan, and Phillip Isola. Learning vision from mod- els rivals learning vision from data. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 15887–15898, 2024. 2

    2024

  46. [54]

    Anti-dreambooth: Pro- tecting users from personalized text-to-image synthesis

    Thanh Van Le, Hao Phung, Thuan Hoang Nguyen, Quan Dao, Ngoc N Tran, and Anh Tran. Anti-dreambooth: Pro- tecting users from personalized text-to-image synthesis. In ICCV, 2023. 2

    2023

  47. [55]

    Sun database: Large-scale scene recognition from abbey to zoo

    Jianxiong Xiao, James Hays, Krista A Ehinger, Aude Oliva, and Antonio Torralba. Sun database: Large-scale scene recognition from abbey to zoo. In CVPR, pages 3485–3492. IEEE, 2010. 6

    2010

  48. [56]

    Synaug: Exploiting synthetic data for data imbalance problems

    Moon Ye-Bin, Nam Hyeon-Woo, Wonseok Choi, Nayeong Kim, Suha Kwak, and Tae-Hyun Oh. Synaug: Exploiting synthetic data for data imbalance problems. arXiv preprint arXiv:2308.00994, 2023. 2

    2023 arXiv

  49. [57]

    Controlled training data generation with diffusion mod- els

    Teresa Yeo, Andrei Atanov, Harold Benoit, Aleksandr Alek- seev, Ruchira Ray, Pooya Esmaeil Akhoondi, and Amir Za- mir. Controlled training data generation with diffusion mod- els. arXiv preprint arXiv:2403.15309, 2024. 2

    2024 arXiv

  50. [58]

    Diffusion models and semi-supervised learners benefit mutually with few labels

    Zebin You, Yong Zhong, Fan Bao, Jiacheng Sun, Chongx- uan Li, and Jun Zhu. Diffusion models and semi-supervised learners benefit mutually with few labels. NeurIPS, 2023. 2

    2023

  51. [59]

    Diversify, don’t fine-tune: Scaling up visual recognition training with synthetic images

    Zhuoran Yu, Chenchen Zhu, Sean Culatana, Raghuraman Krishnamoorthi, Fanyi Xiao, and Yong Jae Lee. Diversify, don’t fine-tune: Scaling up visual recognition training with synthetic images. arXiv preprint arXiv:2312.02253, 2023. 1, 2, 4

    2023 arXiv

  52. [60]

    Real-fake: Effective training data synthesis through distribution matching

    Jianhao Yuan, Jie Zhang, Shuyang Sun, Philip Torr, and Bo Zhao. Real-fake: Effective training data synthesis through distribution matching. In ICLR, 2024. 2

    2024

  53. [61]

    Diffmorpher: Unleashing the capability of diffu- sion models for image morphing

    Kaiwen Zhang, Yifan Zhou, Xudong Xu, Bo Dai, and Xin- gang Pan. Diffmorpher: Unleashing the capability of diffu- sion models for image morphing. In CVPR, 2024. 2

    2024

  54. [62]

    Tip- adapter: Training-free adaption of clip for few-shot classi- fication

    Renrui Zhang, Wei Zhang, Rongyao Fang, Peng Gao, Kun- chang Li, Jifeng Dai, Yu Qiao, and Hongsheng Li. Tip- adapter: Training-free adaption of clip for few-shot classi- fication. In ECCV, 2022. 6

    2022

  55. [63]

    Prompt, generate, then cache: Cascade of foundation models makes strong few-shot learners

    Renrui Zhang, Xiangfei Hu, Bohao Li, Siyuan Huang, Han- qiu Deng, Yu Qiao, Peng Gao, and Hongsheng Li. Prompt, generate, then cache: Cascade of foundation models makes strong few-shot learners. In CVPR, 2023. 2

    2023

  56. [64]

    Toward understanding generative data augmentation.NeurIPS, 2023

    Chenyu Zheng, Guoqiang Wu, and Chongxuan Li. Toward understanding generative data augmentation.NeurIPS, 2023. 2

    2023

  57. [65]

    Learning to prompt for vision-language models

    Kaiyang Zhou, Jingkang Yang, Chen Change Loy, and Ziwei Liu. Learning to prompt for vision-language models. IJCV,

  58. [66]

    Jeep Wrangler SUV 2012

    Yongchao Zhou, Hshmat Sahak, and Jimmy Ba. Training on thin air: Improve image classification with generated data. arXiv preprint arXiv:2305.15316, 2023. 2 11 Supplementary Material for LoFT: LoRA-Fused Training Dataset Generation with Few-shot Guidance A. Implementation detai...

    2023 arXiv

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.