pith. sign in

arxiv: 2605.21272 · v1 · pith:R7XZ2OCYnew · submitted 2026-05-20 · 💻 cs.CV · cs.AI

MONET: A Massive, Open, Non-redundant and Enriched Text-to-image dataset

Benjamin Aubin , Gonzalo I\~naki Quintana , Onur Tasar , Sanjeev Sreetharan , Urszula Czerwinska , Damien Henry , Cl\'ement Chadebec This is my paper

Pith reviewed 2026-05-21 05:17 UTC · model grok-4.3

classification 💻 cs.CV cs.AI
keywords text-to-image datasetdiffusion modeldata curationimage captioningdeduplicationopen datasetlatent diffusion
0
0 comments X

The pith

A 104.9 million pair open dataset, built from 2.9 billion raw images through filtering and re-captioning, supports training of a competitive 4-billion-parameter text-to-image model.

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

The paper introduces MONET as an Apache 2.0 dataset of roughly 104.9 million image-text pairs assembled from heterogeneous open sources. Successive processing steps remove unsafe or duplicate content and generate both short and long captions using several vision-language models, with additional synthetic samples added for coverage. The authors then train a 4B-parameter latent diffusion model using only this dataset and obtain competitive results on standard text-to-image benchmarks. This outcome indicates that careful curation of public data can produce a resource sufficient for large-scale model training. The work thereby reduces dependence on proprietary corpora for reproducible research.

Core claim

Through successive stages of safety filtering, domain-based filtering, exact and near-duplicate removal, and re-captioning with multiple vision-language models covering short to long-form descriptions, and further augmented with synthetically generated samples, we produce an open dataset of approximately 104.9 million image-text pairs that, when used exclusively to train a 4B-parameter latent diffusion model, yields competitive GenEval and DPG scores.

What carries the argument

The multi-stage curation pipeline of safety filtering, domain filtering, exact and near-duplicate removal, and re-captioning by several vision-language models that converts 2.9 billion raw pairs into a non-redundant, enriched collection of 104.9 million pairs.

If this is right

  • Any researcher can download the full dataset under an open license and reproduce large-scale text-to-image training without access to private corpora.
  • Pre-computed embeddings and annotations shipped with each image shorten the time needed for downstream experiments and fine-tuning.
  • The same curation sequence can be reapplied to future waves of public image data to keep the dataset current.
  • Competitive benchmark scores achieved with an exclusively open dataset remove a practical obstacle to community-driven model development.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Similar staged curation could be tested on video or audio generation tasks to check whether the same quality gains appear in other modalities.
  • The dataset's re-captioning step might be studied to measure how much caption style affects final model behavior on specific prompt types.
  • Community extensions could add geographic or cultural tags to the existing annotations to test for improved fairness in generated outputs.

Load-bearing premise

The filtering, deduplication, and multi-model re-captioning steps preserve enough diversity and descriptive quality to train a high-performing model without major loss of information or introduction of new biases.

What would settle it

A direct side-by-side training run of the same 4B-parameter model on a version of the data that skips one or more of the described filtering or re-captioning stages, followed by measurement of the resulting drop in GenEval or DPG scores.

Figures

Figures reproduced from arXiv: 2605.21272 by Benjamin Aubin, Cl\'ement Chadebec, Damien Henry, Gonzalo I\~naki Quintana, Onur Tasar, Sanjeev Sreetharan, Urszula Czerwinska.

Figure 1
Figure 1. Figure 1: An impressionist water-lily painting generated at [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Curation pipeline of the MONET. Each stage removes images that fail the corresponding [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Examples of images at different aesthetic scores [ [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: SSCD nearest-neighbor pairs with cosine similarity and pHash Hamming distance [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Examples of synthetic images generated for the MONET dataset using different models. [PITH_FULL_IMAGE:figures/full_fig_p006_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Representative re-captioning example, comparing the [PITH_FULL_IMAGE:figures/full_fig_p007_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Captions and image statistics of MONET. Image resolution is in Megapixels (MP). [PITH_FULL_IMAGE:figures/full_fig_p008_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: MONET dataset distribution: (left) YOLO-based content classification, (middle) CLIP [PITH_FULL_IMAGE:figures/full_fig_p009_8.png] view at source ↗
Figure 10
Figure 10. Figure 10: (Left) Long-CLIP score evolution throughout training with different captioning models [PITH_FULL_IMAGE:figures/full_fig_p009_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Generation from our 4B model trained exclusively on MONET, showcasing its ability to learn complex concepts and a variety of styles at 1024 × 1024 and 2048 × 2048 resolutions. 11 [PITH_FULL_IMAGE:figures/full_fig_p011_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Distribution and examples of watermark scores. [PITH_FULL_IMAGE:figures/full_fig_p022_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Distribution and examples of Jasper NSFW score. No band 5 examples are included. [PITH_FULL_IMAGE:figures/full_fig_p023_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Distribution and examples of Bumble NSFW score. [PITH_FULL_IMAGE:figures/full_fig_p024_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: Histogram in logarithmic scale, with filtering threshold. [PITH_FULL_IMAGE:figures/full_fig_p024_15.png] view at source ↗
Figure 16
Figure 16. Figure 16: Duplicate clusters detected by perceptual hashing at increasing Hamming distance [PITH_FULL_IMAGE:figures/full_fig_p025_16.png] view at source ↗
Figure 17
Figure 17. Figure 17: pHash false positive clusters: each row shows images that pHash assigns a low Hamming [PITH_FULL_IMAGE:figures/full_fig_p026_17.png] view at source ↗
Figure 18
Figure 18. Figure 18: Distribution of SSCD nearest-neighbor maximum cosine similarities. [PITH_FULL_IMAGE:figures/full_fig_p027_18.png] view at source ↗
Figure 19
Figure 19. Figure 19: SSCD threshold sweep. Nearest-neighbor pairs sampled at increasing cosine similarity. [PITH_FULL_IMAGE:figures/full_fig_p027_19.png] view at source ↗
Figure 20
Figure 20. Figure 20: Examples of near-duplicate clusters detected by SSCD. Each row shows one cluster; the [PITH_FULL_IMAGE:figures/full_fig_p029_20.png] view at source ↗
Figure 21
Figure 21. Figure 21: SSCD false positives on template-based content. [PITH_FULL_IMAGE:figures/full_fig_p030_21.png] view at source ↗
Figure 22
Figure 22. Figure 22: Re-captioning example (1/5) 32 [PITH_FULL_IMAGE:figures/full_fig_p032_22.png] view at source ↗
Figure 23
Figure 23. Figure 23: Re-captioning example (2/5) 33 [PITH_FULL_IMAGE:figures/full_fig_p033_23.png] view at source ↗
Figure 24
Figure 24. Figure 24: Re-captioning example (3/5) 34 [PITH_FULL_IMAGE:figures/full_fig_p034_24.png] view at source ↗
Figure 25
Figure 25. Figure 25: Re-captioning example (4/5) 35 [PITH_FULL_IMAGE:figures/full_fig_p035_25.png] view at source ↗
Figure 26
Figure 26. Figure 26: Re-captioning example (4/5) 36 [PITH_FULL_IMAGE:figures/full_fig_p036_26.png] view at source ↗
Figure 27
Figure 27. Figure 27: Human Elo scores aggregated from the pairwise voting study, plotted against the cosine [PITH_FULL_IMAGE:figures/full_fig_p037_27.png] view at source ↗
Figure 28
Figure 28. Figure 28: Hierarchical image content distribution (YOLO). [PITH_FULL_IMAGE:figures/full_fig_p038_28.png] view at source ↗
Figure 29
Figure 29. Figure 29: Hierarchical image content distribution (CLIP). [PITH_FULL_IMAGE:figures/full_fig_p039_29.png] view at source ↗
Figure 30
Figure 30. Figure 30: Examples of top 5 image content classification using CLIP with their similirarity scores. [PITH_FULL_IMAGE:figures/full_fig_p039_30.png] view at source ↗
Figure 31
Figure 31. Figure 31: Examples of top-5 image content classifications using CLIP, including some incorrect or [PITH_FULL_IMAGE:figures/full_fig_p040_31.png] view at source ↗
Figure 32
Figure 32. Figure 32: shows the detailed distribution of image styles, as in [PITH_FULL_IMAGE:figures/full_fig_p042_32.png] view at source ↗
Figure 33
Figure 33. Figure 33: Example images per picture-style label sampled from the audit preview. [PITH_FULL_IMAGE:figures/full_fig_p043_33.png] view at source ↗
Figure 34
Figure 34. Figure 34: Generation from our 4B model (1024 × 1024) showcasing its ability to generate high resolution images thanks to the MONET Dataset. 47 [PITH_FULL_IMAGE:figures/full_fig_p047_34.png] view at source ↗
Figure 35
Figure 35. Figure 35: Generation from our 4B model (1024 × 1024) showcasing its ability to generate images with different styles thanks to the MONET Dataset. 48 [PITH_FULL_IMAGE:figures/full_fig_p048_35.png] view at source ↗
Figure 36
Figure 36. Figure 36: 2048 × 2048 generation from our 4B model. 49 [PITH_FULL_IMAGE:figures/full_fig_p049_36.png] view at source ↗
Figure 37
Figure 37. Figure 37: 2048 × 2048 generation from our 4B model. 50 [PITH_FULL_IMAGE:figures/full_fig_p050_37.png] view at source ↗
Figure 38
Figure 38. Figure 38: Aggregate distributions from the VLM-based ethics audit over twelve dimensions: cultural [PITH_FULL_IMAGE:figures/full_fig_p051_38.png] view at source ↗
read the original abstract

Training large text-to-image models requires high-quality, curated datasets with diverse content and detailed captions. Yet the cost and complexity of collecting, filtering, deduplicating, and re-captioning such corpora at scale hinders open and reproducible research in the field. We introduce MONET, an open Apache 2.0 dataset of approx. 104.9M image--text pairs collected from 2.9B raw pairs across heterogeneous open sources through successive stages of safety filtering, domain-based filtering, exact and near-duplicate removal, and re-captioning with multiple vision-language models covering short to long-form descriptions, and further augmented with synthetically generated samples. Each image is shipped with pre-computed embeddings and annotations to accelerate downstream use. To validate the effectiveness of MONET, we train a 4B-parameter latent diffusion model exclusively on it and reach competitive GenEval and DPG scores, demonstrating that our dataset lowers the barrier to large-scale, reproducible text-to-image research.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript introduces MONET, an open Apache 2.0 dataset of approximately 104.9 million image-text pairs derived from 2.9 billion raw pairs collected from heterogeneous open sources. Construction proceeds through successive stages of safety filtering, domain-based filtering, exact and near-duplicate removal, re-captioning with multiple vision-language models (short to long-form), and augmentation with synthetic samples. Each image is accompanied by pre-computed embeddings and annotations. Effectiveness is validated by training a 4B-parameter latent diffusion model exclusively on the final dataset and reporting competitive GenEval and DPG scores, with the goal of lowering barriers to large-scale reproducible text-to-image research.

Significance. If the curation pipeline demonstrably yields a high-quality, diverse corpus without substantial information loss or introduced biases, MONET would constitute a valuable open resource for the computer vision community. It would enable reproducible training of large text-to-image models at scale and reduce dependence on closed datasets. The inclusion of embeddings and annotations further increases practical utility for downstream tasks.

major comments (2)
  1. [Abstract / Validation section] Abstract and validation experiment: the manuscript states that a 4B-parameter latent diffusion model trained exclusively on MONET reaches competitive GenEval and DPG scores, yet supplies no numerical benchmark values, no baseline comparisons against models trained on other open datasets, and no training hyperparameters or ablation results. This leaves the central effectiveness claim without visible quantitative support.
  2. [Dataset construction] Dataset construction pipeline: no ablations, retained-concept metrics, or bias measurements are reported that isolate the contribution of safety filtering, domain filtering, deduplication, or multi-VLM re-captioning. Without such controls it is impossible to verify that the successive stages improve quality or diversity rather than merely preserving scale from the 2.9 B starting corpus plus synthetic augmentation.
minor comments (2)
  1. [Dataset construction] A table summarizing the exact number of pairs retained after each filtering and re-captioning stage would improve transparency and allow readers to assess information loss quantitatively.
  2. [Dataset construction] Clarify whether the synthetic augmentation samples are generated from the filtered MONET captions or from an external model, and report their proportion in the final 104.9 M set.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive and detailed review. We address each major comment below with clarifications on the current manuscript content and indicate where revisions will be made to strengthen the presentation.

read point-by-point responses

  1. Referee: [Abstract / Validation section] Abstract and validation experiment: the manuscript states that a 4B-parameter latent diffusion model trained exclusively on MONET reaches competitive GenEval and DPG scores, yet supplies no numerical benchmark values, no baseline comparisons against models trained on other open datasets, and no training hyperparameters or ablation results. This leaves the central effectiveness claim without visible quantitative support.

    Authors: We acknowledge that the abstract presents only a high-level statement of competitive performance. The full manuscript reports the specific GenEval and DPG scores achieved by the 4B model in the validation section, along with basic training details. However, we agree that explicit numerical comparisons to models trained on other open datasets (e.g., LAION subsets) and a fuller set of hyperparameters would provide stronger quantitative support. We will revise the validation section and add a comparison table in the next version. revision: yes

  2. Referee: [Dataset construction] Dataset construction pipeline: no ablations, retained-concept metrics, or bias measurements are reported that isolate the contribution of safety filtering, domain filtering, deduplication, or multi-VLM re-captioning. Without such controls it is impossible to verify that the successive stages improve quality or diversity rather than merely preserving scale from the 2.9 B starting corpus plus synthetic augmentation.

    Authors: We agree that isolating the contribution of each stage via ablations would be valuable. At the scale of the initial 2.9 billion pairs, however, training separate models on every intermediate dataset is computationally prohibitive. The manuscript does include overall statistics on size reduction after deduplication and filtering, as well as diversity indicators. We will expand the construction section with additional retained-concept and bias metrics drawn from our processing logs in the revision. revision: partial

Circularity Check

0 steps flagged

No circularity; validation uses independent external benchmarks

full rationale

The paper describes a curation pipeline of safety filtering, domain filtering, deduplication and multi-VLM re-captioning applied to 2.9B raw pairs, then trains a 4B LDM exclusively on the resulting 104.9M pairs and reports competitive scores on the standard GenEval and DPG benchmarks. These benchmarks are defined and computed outside the curation process and are not fitted to or defined by any pipeline parameter, so the reported performance constitutes an independent empirical check rather than a quantity that reduces to the inputs by construction. No self-citations, self-definitional loops, fitted predictions, or imported uniqueness theorems appear in the derivation chain.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the unverified assumption that the described filtering and recaptioning pipeline yields higher-quality training data than raw or existing open corpora; this is treated as a domain assumption without independent evidence or metrics supplied in the abstract.

axioms (1)
  • domain assumption Recaptioning with multiple vision-language models produces accurate, diverse, and unbiased descriptions that improve downstream model performance.
    This premise underpins the enrichment step but receives no quantitative validation in the provided abstract.

pith-pipeline@v0.9.0 · 5735 in / 1381 out tokens · 71746 ms · 2026-05-21T05:17:41.732293+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

171 extracted references · 171 canonical work pages · 30 internal anchors

  1. [1]

    Fine-tuned vision transformer (vit) for nsfw image classification

    Falcon AI. Fine-tuned vision transformer (vit) for nsfw image classification. https://huggingface. co/Falconsai/nsfw_image_detection, 2024. Accessed: 2026-04-16

    work page 2024

  2. [2]

    Qwen-VL: A Versatile Vision-Language Model for Understanding, Localization, Text Reading, and Beyond

    Jinze Bai, Shuai Bai, Shusheng Yang, Shijie Wang, Sinan Tan, Peng Wang, Junyang Lin, Chang Zhou, and Jingren Zhou. Qwen-vl: A versatile vision-language model for understanding, localization, text reading, and beyond.arXiv preprint arXiv:2308.12966, 2023

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  3. [3]

    Improving image generation with better captions.Computer Science

    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

    work page 2023

  4. [4]

    Language models are few-shot learners

    Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al. Language models are few-shot learners. Advances in neural information processing systems, 33:1877–1901, 2020

    work page 1901

  5. [5]

    Bumble’s private detector model

    Bumble-Tech. Bumble’s private detector model. https://github.com/bumble-ai/nsfw-image-d etection, 2024. Accessed: 2026-04-16

    work page 2024

  6. [6]

    Coyo-700m: Image-text pair dataset.https://github.com/kakaobrain/coyo-dataset, 2022

    Minwoo Byeon, Beomhee Park, Haecheon Kim, Sungjun Lee, Woonhyuk Baek, and Saehoon Kim. Coyo-700m: Image-text pair dataset.https://github.com/kakaobrain/coyo-dataset, 2022

    work page 2022

  7. [7]

    HiDream-I1: A High-Efficient Image Generative Foundation Model with Sparse Diffusion Transformer

    Qi Cai, Jingwen Chen, Yang Chen, Yehao Li, Fuchen Long, Yingwei Pan, Zhaofan Qiu, Yiheng Zhang, Fengbin Gao, Peihan Xu, et al. Hidream-i1: A high-efficient image generative foundation model with sparse diffusion transformer.arXiv preprint arXiv:2505.22705, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  8. [8]

    Extracting training data from diffusion models

    Nicolas Carlini, Jamie Hayes, Milad Nasr, Matthew Jagielski, Vikash Sehwag, Florian Tramer, Borja Balle, Daphne Ippolito, and Eric Wallace. Extracting training data from diffusion models. In32nd USENIX security symposium (USENIX Security 23), pages 5253–5270, 2023

    work page 2023

  9. [9]

    Conceptual 12M: Pushing web-scale image-text pre-training to recognize long-tail visual concepts

    Soravit Changpinyo, Piyush Sharma, Nan Ding, and Radu Soricut. Conceptual 12M: Pushing web-scale image-text pre-training to recognize long-tail visual concepts. InCVPR, 2021

    work page 2021

  10. [10]

    Pixart- α: Fast training of diffusion transformer for photorealistic text-to-image synthesis

    Junsong Chen, YU Jincheng, GE Chongjian, Lewei Yao, Enze Xie, Zhongdao Wang, James Kwok, Ping Luo, Huchuan Lu, and Zhenguo Li. Pixart- α: Fast training of diffusion transformer for photorealistic text-to-image synthesis. InThe Twelfth International Conference on Learning Representations, 2023

    work page 2023

  11. [11]

    Pixart- σ: Weak-to-strong training of diffusion transformer for 4k text-to-image generation.arXiv preprint arXiv:2403.04692, 2024

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

    work page arXiv 2024

  12. [12]

    Deep compression autoencoder for efficient high-resolution diffusion models

    Junyu Chen, Han Cai, Junsong Chen, Enze Xie, Shang Yang, Haotian Tang, Muyang Li, and Song Han. Deep compression autoencoder for efficient high-resolution diffusion models. InThe Thirteenth International Conference on Learning Representations, 2025

    work page 2025

  13. [13]

    Sharegpt4v: Improving large multi-modal models with better captions

    Lin Chen, Jinsong Li, Xiaoyi Dong, Pan Zhang, Conghui He, Jiaqi Wang, Feng Zhao, and Dahua Lin. Sharegpt4v: Improving large multi-modal models with better captions. InEuropean Conference on Computer Vision, pages 370–387. Springer, 2024

    work page 2024

  14. [14]

    Janus-Pro: Unified Multimodal Understanding and Generation with Data and Model Scaling

    Xiaokang Chen, Zhiyu Wu, Xingchao Liu, Zizheng Pan, Wen Liu, Zhenda Xie, Xingkai Yu, and Chong Ruan. Janus-pro: Unified multimodal understanding and generation with data and model scaling.arXiv preprint arXiv:2501.17811, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  15. [15]

    Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities

    Gheorghe Comanici, Eric Bieber, Mike Schaekermann, Ice Pasupat, Noveen Sachdeva, Inderjit Dhillon, Marcel Blistein, Ori Ram, Dan Zhang, Evan Rosen, et al. Gemini 2.5: Pushing the frontier with advanced reasoning, multimodality, long context, and next generation agentic capabilities.arXiv preprint arXiv:2507.06261, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  16. [16]

    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. In2009 IEEE conference on computer vision and pattern recognition, pages 248–255. Ieee, 2009

    work page 2009

  17. [17]

    Redcaps: Web-curated image-text data created by the people, for the people

    Karan Desai, Gaurav Kaul, Zubin Trivadi Aysola, and Justin Johnson. Redcaps: Web-curated image-text data created by the people, for the people. InThirty-fifth Conference on Neural Information Processing Systems Datasets and Benchmarks Track (Round 1), 2021. URL https://openreview.net/forum?i d=VjJxBi1p9zh. 13

    work page 2021

  18. [18]

    The Faiss library

    Matthijs Douze, Alexandr Guzhva, Chengqi Deng, Jeff Johnson, Gergely Szilvasy, Pierre-Emmanuel Mazaré, Maria Lomeli, Lucas Hosseini, and Hervé Jegou. The faiss library.arXiv preprint arXiv:2401.08281, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  19. [19]

    Scaling Rectified Flow Transformers for High-Resolution Image Synthesis

    Patrick Esser, Sumith Kulal, Andreas Blattmann, Rahim Entezari, Jonas Müller, Harry Saini, Yam Levi, Dominik Lorenz, Axel Sauer, Frederic Boesel, et al. Scaling rectified flow transformers for high-resolution image synthesis.arXiv preprint arXiv:2403.03206, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  20. [20]

    Data filtering networks

    Alex Fang, Albin Madappally Jose, Amit Jain, Ludwig Schmidt, Alexander T Toshev, and Vaishaal Shankar. Data filtering networks. InThe Twelfth International Conference on Learning Representations,

    work page

  21. [21]

    URLhttps://openreview.net/forum?id=KAk6ngZ09F

    work page

  22. [22]

    The validity and practicality of sun-reactive skin types I through VI.Archives of Dermatology, 124(6):869–871, 1988

    Thomas B Fitzpatrick. The validity and practicality of sun-reactive skin types I through VI.Archives of Dermatology, 124(6):869–871, 1988

    work page 1988

  23. [23]

    Datacomp: In search of the next generation of multimodal datasets.Advances in Neural Information Processing Systems, 36:27092–27112, 2023

    Samir Yitzhak Gadre, Gabriel Ilharco, Alex Fang, Jonathan Hayase, Georgios Smyrnis, Thao Nguyen, Ryan Marten, Mitchell Wortsman, Dhruba Ghosh, Jieyu Zhang, et al. Datacomp: In search of the next generation of multimodal datasets.Advances in Neural Information Processing Systems, 36:27092–27112, 2023

    work page 2023

  24. [24]

    Seedream 3.0 Technical Report

    Yu Gao, Lixue Gong, Qiushan Guo, Xiaoxia Hou, Zhichao Lai, Fanshi Li, Liang Li, Xiaochen Lian, Chao Liao, Liyang Liu, et al. Seedream 3.0 technical report.arXiv preprint arXiv:2504.11346, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  25. [25]

    Datasheets for datasets.Communications of the ACM, 64(12):86–92, 2021

    Timnit Gebru, Jamie Morgenstern, Briana Vecchione, Jennifer Wortman Vaughan, Hanna Wallach, Hal Daumé III, and Kate Crawford. Datasheets for datasets.Communications of the ACM, 64(12):86–92, 2021

    work page 2021

  26. [26]

    Geneval: An object-focused framework for evaluating text-to-image alignment.Advances in Neural Information Processing Systems, 36:52132– 52152, 2023

    Dhruba Ghosh, Hannaneh Hajishirzi, and Ludwig Schmidt. Geneval: An object-focused framework for evaluating text-to-image alignment.Advances in Neural Information Processing Systems, 36:52132– 52152, 2023

    work page 2023

  27. [27]

    Commoncanvas: An open diffusion model trained with creative-commons images.arXiv preprint arXiv:2310.16825, 2023

    Aaron Gokaslan, A Feder Cooper, Jasmine Collins, Landan Seguin, Austin Jacobson, Mihir Patel, Jonathan Frankle, Cory Stephenson, and V olodymyr Kuleshov. Commoncanvas: An open diffusion model trained with creative-commons images.arXiv preprint arXiv:2310.16825, 2023

    work page arXiv 2023

  28. [28]

    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. InAdvances in Neural Information Processing Systems, pages 2672–2680, 2014

    work page 2014

  29. [29]

    The Llama 3 Herd of Models

    Aaron Grattafiori, Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Alex Vaughan, et al. The llama 3 herd of models. arXiv preprint arXiv:2407.21783, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  30. [30]

    On memorization in diffusion models.Transactions on Machine Learning Research, 2025

    Xiangming Gu, Chao Du, Tianyu Pang, Chongxuan Li, Min Lin, and Ye Wang. On memorization in diffusion models.Transactions on Machine Learning Research, 2025. ISSN 2835-8856. URL https://openreview.net/forum?id=D3DBqvSDbj

    work page 2025

  31. [31]

    Gans trained by a two time-scale update rule converge to a local nash equilibrium.Advances in neural information processing systems, 30, 2017

    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.Advances in neural information processing systems, 30, 2017

    work page 2017

  32. [32]

    Classifier-free diffusion guidance

    Jonathan Ho and Tim Salimans. Classifier-free diffusion guidance. InNeurIPS 2021 Workshop on Deep Generative Models and Downstream Applications, 2021

    work page 2021

  33. [33]

    Denoising diffusion probabilistic models.Advances in neural information processing systems, 33:6840–6851, 2020

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising diffusion probabilistic models.Advances in neural information processing systems, 33:6840–6851, 2020

    work page 2020

  34. [34]

    Imagen Video: High Definition Video Generation with Diffusion Models

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

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  35. [35]

    Training Compute-Optimal Large Language Models

    Jordan Hoffmann, Sebastian Borgeaud, Arthur Mensch, Elena Buchatskaya, Trevor Cai, Eliza Rutherford, DDL Casas, Lisa Anne Hendricks, Johannes Welbl, Aidan Clark, et al. Training compute-optimal large language models.arXiv preprint arXiv:2203.15556, 10, 2022

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  36. [36]

    CogVLM2: Visual Language Models for Image and Video Understanding

    Wenyi Hong, Weihan Wang, Ming Ding, Wenmeng Yu, Qingsong Lv, Yan Wang, Yean Cheng, Shiyu Huang, Junhui Ji, Zhao Xue, et al. Cogvlm2: Visual language models for image and video understanding. arXiv preprint arXiv:2408.16500, 2024. 14

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  37. [37]

    Scaling up vision-language pre-training for image captioning

    Xiaowei Hu, Zhe Gan, Jianfeng Wang, Zhengyuan Yang, Zicheng Liu, Yumao Lu, and Lijuan Wang. Scaling up vision-language pre-training for image captioning. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 17980–17989, 2022

    work page 2022

  38. [38]

    ELLA: Equip Diffusion Models with LLM for Enhanced Semantic Alignment

    Xiwei Hu, Rui Wang, Yixiao Fang, Bin Fu, Pei Cheng, and Gang Yu. Ella: Equip diffusion models with llm for enhanced semantic alignment.arXiv preprint arXiv:2403.05135, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  39. [39]

    GPT-4o System Card

    Aaron Hurst, Adam Lerer, Adam P Goucher, Adam Perelman, Aditya Ramesh, Aidan Clark, AJ Ostrow, Akila Welihinda, Alan Hayes, Alec Radford, et al. Gpt-4o system card.arXiv preprint arXiv:2410.21276, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  40. [40]

    Scaling up visual and vision-language representation learning with noisy text supervision

    Chao Jia, Yinfei Yang, Ye Xia, Yi-Ting Chen, Zarana Parekh, Hieu Pham, Quoc Le, Yun-Hsuan Sung, Zhen Li, and Tom Duerig. Scaling up visual and vision-language representation learning with noisy text supervision. InInternational conference on machine learning, pages 4904–4916. PMLR, 2021

    work page 2021

  41. [41]

    Mixtral of Experts

    Albert Q Jiang, Alexandre Sablayrolles, Antoine Roux, Arthur Mensch, Blanche Savary, Chris Bamford, Devendra Singh Chaplot, Diego de las Casas, Emma Bou Hanna, Florian Bressand, et al. Mixtral of experts.arXiv preprint arXiv:2401.04088, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  42. [42]

    Ultralytics YOLOv8

    Glenn Jocher, Ayush Chaurasia, and Jing Qiu. Ultralytics YOLOv8. https://github.com/ultraly tics/ultralytics, 2023

    work page 2023

  43. [43]

    Improved FLUX prompts dataset

    k-mktr. Improved FLUX prompts dataset. https://huggingface.co/datasets/k-mktr/improv ed-flux-prompts, 2024. Accessed: 2026-05-05

    work page 2024

  44. [44]

    Deduplicating training data mitigates privacy risks in language models

    Nikhil Kandpal, Eric Wallace, and Colin Raffel. Deduplicating training data mitigates privacy risks in language models. InInternational Conference on Machine Learning, pages 10697–10707. PMLR, 2022

    work page 2022

  45. [45]

    Scaling up gans for text-to-image synthesis

    Minguk Kang, Jun-Yan Zhu, Richard Zhang, Jaesik Park, Eli Shechtman, Sylvain Paris, and Taesung Park. Scaling up gans for text-to-image synthesis. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 10124–10134, 2023

    work page 2023

  46. [46]

    Adam: A Method for Stochastic Optimization

    Diederik P Kingma and Jimmy Ba. Adam: A method for stochastic optimization.arXiv preprint arXiv:1412.6980, 2014

    work page internal anchor Pith review Pith/arXiv arXiv 2014

  47. [47]

    jina-clip-v2: Multilingual multimodal embeddings for text and images.arXiv preprint arXiv:2412.08802, 2024

    Andreas Koukounas, Georgios Mastrapas, Michael Günther, Bo Wang, Scott Martens, Isabelle Mohr, Saba Sturua, Mohammad Kalim Akram, Joan Fontanals Martínez, Saahil Ognawala, et al. jina-clip-v2: Multilingual multimodal embeddings for text and images.arXiv preprint arXiv:2412.08802, 2024

    work page arXiv 2024

  48. [48]

    Visual genome: Connecting language and vision using crowdsourced dense image annotations.International journal of computer vision, 123(1):32–73, 2017

    Ranjay Krishna, Yuke Zhu, Oliver Groth, Justin Johnson, Kenji Hata, Joshua Kravitz, Stephanie Chen, Yannis Kalantidis, Li-Jia Li, David A Shamma, et al. Visual genome: Connecting language and vision using crowdsourced dense image annotations.International journal of computer vision, 123(1):32–73, 2017

    work page 2017

  49. [49]

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

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

    work page 2024

  50. [50]

    FLUX.2: Frontier Visual Intelligence.https://bfl.ai/blog/flux-2, 2025

    Black Forest Labs. FLUX.2: Frontier Visual Intelligence.https://bfl.ai/blog/flux-2, 2025

    work page 2025

  51. [51]

    Releasing re-laion-5b: transparent iteration on laion-5b with additional safety fixes

    LAION. Releasing re-laion-5b: transparent iteration on laion-5b with additional safety fixes. https: //laion.ai/blog/relaion-5b/, 2024. Accessed: 30 aug, 2024

    work page 2024

  52. [52]

    Aesthetic predictor

    LAION-AI. Aesthetic predictor. https://github.com/christophschuhmann/improved-aesthet ic-predictor, 2022. Accessed: 2026-04-03

    work page 2022

  53. [53]

    Deduplicating training data makes language models better

    Katherine Lee, Daphne Ippolito, Andrew Nystrom, Chiyuan Zhang, Douglas Eck, Chris Callison-Burch, and Nicholas Carlini. Deduplicating training data makes language models better. InProceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 8424–8445, 2022

    work page 2022

  54. [54]

    Playground v2.5: Three Insights towards Enhancing Aesthetic Quality in Text-to-Image Generation

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

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  55. [55]

    Blip-2: Bootstrapping language-image pre- training with frozen image encoders and large language models

    Junnan Li, Dongxu Li, Silvio Savarese, and Steven Hoi. Blip-2: Bootstrapping language-image pre- training with frozen image encoders and large language models. InInternational conference on machine learning, pages 19730–19742. PMLR, 2023

    work page 2023

  56. [56]

    vunderstanding and generation with masked discrete diffusion.arXiv preprint arXiv:2603.06577, 2026

    Lijiang Li, Zuwei Long, Yunhang Shen, Heting Gao, Haoyu Cao, Xing Sun, Caifeng Shan, Ran He, and Chaoyou Fu. vunderstanding and generation with masked discrete diffusion.arXiv preprint arXiv:2603.06577, 2026. 15

    work page arXiv 2026

  57. [57]

    Hunyuan-DiT: A Powerful Multi-Resolution Diffusion Transformer with Fine-Grained Chinese Understanding

    Zhimin Li, Jianwei Zhang, Qin Lin, Jiangfeng Xiong, Yanxin Long, Xinchi Deng, Yingfang Zhang, Xingchao Liu, Minbin Huang, Zedong Xiao, et al. Hunyuan-dit: A powerful multi-resolution diffusion transformer with fine-grained chinese understanding.arXiv preprint arXiv:2405.08748, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  58. [58]

    Microsoft coco: Common objects in context

    Tsung-Yi Lin, Michael Maire, Serge Belongie, James Hays, Pietro Perona, Deva Ramanan, Piotr Dollár, and C Lawrence Zitnick. Microsoft coco: Common objects in context. InComputer Vision–ECCV 2014: 13th European Conference, Zurich, Switzerland, September 6-12, 2014, Proceedings, Part V 13, pages 740–755. Springer, 2014

    work page 2014

  59. [59]

    Flow matching for generative modeling

    Yaron Lipman, Ricky TQ Chen, Heli Ben-Hamu, Maximilian Nickel, and Matthew Le. Flow matching for generative modeling. InThe Eleventh International Conference on Learning Representations, 2023

    work page 2023

  60. [60]

    DeepSeek-V3 Technical Report

    Aixin Liu, Bei Feng, Bing Xue, Bingxuan Wang, Bochao Wu, Chengda Lu, Chenggang Zhao, Chengqi Deng, Chenyu Zhang, Chong Ruan, et al. Deepseek-v3 technical report.arXiv preprint arXiv:2412.19437, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  61. [61]

    Flow straight and fast: Learning to generate and transfer data with rectified flow

    Xingchao Liu, Chengyue Gong, et al. Flow straight and fast: Learning to generate and transfer data with rectified flow. InThe Eleventh International Conference on Learning Representations, 2022

    work page 2022

  62. [62]

    MediaPipe: A Framework for Building Perception Pipelines

    Camillo Lugaresi, Jiuqiang Tang, Hadon Nash, Chris McClanahan, Esha Uboweja, Michael Hays, Fan Zhang, Chuo-Ling Chang, Ming Guang Yong, Juhyun Lee, et al. MediaPipe: A framework for building perception pipelines.arXiv preprint arXiv:1906.08172, 2019

    work page internal anchor Pith review Pith/arXiv arXiv 1906

  63. [63]

    v: Harmonizing autoregression and rectified flow for unified multimodal understanding and generation

    Yiyang Ma, Xingchao Liu, Xiaokang Chen, Wen Liu, Chengyue Wu, Zhiyu Wu, Zizheng Pan, Zhenda Xie, Haowei Zhang, Xingkai Yu, et al. v: Harmonizing autoregression and rectified flow for unified multimodal understanding and generation. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 7739–7751, 2025

    work page 2025

  64. [64]

    Gpt-image-1, 2025

    OpenAI. Gpt-image-1, 2025. URL https://openai.com/zh-Hans-CN/index/introducing-4 o-image-generation/

    work page 2025

  65. [65]

    Dinov2: Learning robust visual features without supervision.Transactions on Machine Learning Research, 2024

    Maxime Oquab, Timothée Darcet, Théo Moutakanni, Huy V o, Marc Szafraniec, Vasil Khalidov, Pierre Fernandez, Daniel Haziza, Francisco Massa, Alaaeldin El-Nouby, et al. Dinov2: Learning robust visual features without supervision.Transactions on Machine Learning Research, 2024

    work page 2024

  66. [66]

    Scalable diffusion models with transformers

    William Peebles and Saining Xie. Scalable diffusion models with transformers. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 4195–4205, 2023

    work page 2023

  67. [67]

    A self- supervised descriptor for image copy detection.Proc

    Ed Pizzi, Sreya Dutta Roy, Sugosh Nagavara Ravindra, Priya Goyal, and Matthijs Douze. A self- supervised descriptor for image copy detection.Proc. CVPR, 2022

    work page 2022

  68. [68]

    SSCD: A self-supervised descriptor for image copy detection – code and pretrained models

    Ed Pizzi, Sreya Dutta Roy, Sugosh Nagavara Ravindra, Priya Goyal, and Matthijs Douze. SSCD: A self-supervised descriptor for image copy detection – code and pretrained models. https://github.c om/facebookresearch/sscd-copy-detection, 2022. Accessed: 2026-04-23

    work page 2022

  69. [69]

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

    Dustin Podell, Zion English, Kyle Lacey, Andreas Blattmann, Tim Dockhorn, Jonas Müller, Joe Penna, and Robin Rombach. Sdxl: Improving latent diffusion models for high-resolution image synthesis. In The Twelfth International Conference on Learning Representations, 2023

    work page 2023

  70. [70]

    Lumina-image 2.0: A unified and efficient image generative framework

    Qi Qin, Le Zhuo, Yi Xin, Ruoyi Du, Zhen Li, Bin Fu, Yiting Lu, Xinyue Li, Dongyang Liu, Xiangyang Zhu, et al. Lumina-image 2.0: A unified and efficient image generative framework. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 20031–20042, 2025

    work page 2025

  71. [71]

    Learning transferable visual models from natural language supervision

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning transferable visual models from natural language supervision. InInternational conference on machine learning, pages 8748–8763. PMLR, 2021

    work page 2021

  72. [72]

    Exploring the limits of transfer learning with a unified text-to-text transformer

    Colin Raffel, Noam Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, and Peter J Liu. Exploring the limits of transfer learning with a unified text-to-text transformer. Journal of machine learning research, 21(140):1–67, 2020

    work page 2020

  73. [73]

    Zero-shot text-to-image generation

    Aditya Ramesh, Mikhail Pavlov, Gabriel Goh, Scott Gray, Chelsea V oss, Alec Radford, Mark Chen, and Ilya Sutskever. Zero-shot text-to-image generation. InInternational conference on machine learning, pages 8821–8831. Pmlr, 2021

    work page 2021

  74. [74]

    Hierarchical Text-Conditional Image Generation with CLIP Latents

    Aditya Ramesh, Prafulla Dhariwal, Alex Nichol, Casey Chu, and Mark Chen. Hierarchical text-conditional image generation with clip latents.arXiv preprint arXiv:2204.06125, 2022. 16

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  75. [75]

    Kandinsky: an improved text-to-image synthesis with image prior and latent diffusion

    Anton Razzhigaev, Arseniy Shakhmatov, Anastasia Maltseva, Vladimir Arkhipkin, Igor Pavlov, Ilya Ryabov, Angelina Kuts, Alexander Panchenko, Andrey Kuznetsov, and Denis Dimitrov. Kandinsky: an improved text-to-image synthesis with image prior and latent diffusion. InProceedings of the 2023 Conference on Empirical Methods in Natural Language Processing: Sys...

    work page 2023

  76. [76]

    You only look once: Unified, real-time object detection

    Joseph Redmon, Santosh Divvala, Ross Girshick, and Ali Farhadi. You only look once: Unified, real-time object detection. InCVPR, pages 779–788, 2016

    work page 2016

  77. [77]

    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. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 10684–10695, 2022

    work page 2022

  78. [78]

    U-net: Convolutional networks for biomedical image segmentation

    Olaf Ronneberger, Philipp Fischer, and Thomas Brox. U-net: Convolutional networks for biomedical image segmentation. InMedical image computing and computer-assisted intervention–MICCAI 2015: 18th international conference, Munich, Germany, October 5-9, 2015, proceedings, part III 18, pages 234–241. Springer, 2015

    work page 2015

  79. [79]

    Photorealistic text-to- image diffusion models with deep language understanding.Advances in neural information processing systems, 35:36479–36494, 2022

    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.Advances in neural information processing systems, 35:36479–36494, 2022

    work page 2022

  80. [80]

    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. InInternational conference on machine learning, pages 30105–30118. PMLR, 2023

    work page 2023

Showing first 80 references.