Pith. sign in

REVIEW 3 major objections 5 minor 1 cited by

Multiverse Through Deepfakes: The MultiFakeVerse Dataset of Person-Centric Visual and Conceptual Manipulations

T0 review · 3 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read The paper introduces MultiFakeVerse, a dataset of 845,286 images in which vision-language models make semantic, person-centric edits to photographs, and reports that current deepfake detectors and humans largely miss these subtle changes.

desk verdict A useful person-centric deepfake dataset whose detection benchmark holds up, but whose localization ground truth is unvalidated and whose finetuning narrative misreads its own table. read the letter →

arxiv 2506.00868 v2 pith:AWWG6G2B submitted 2025-06-01 cs.MM cs.CV

Parul Gupta , Shreya Ghosh , Tom Gedeon , Thanh-Toan Do , Abhinav Dhall This is my paper
classification cs.MMcs.CV
keywords MultiFakeVersedeepfakedetectionperson-centricmanipulationvision-languagemodelsimagesemanticforgerylocalizationbenchmarkdataset
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 MultiFakeVerse, a dataset of 845,286 images built by asking vision-language models to make minimal edits to real photos of people so that the viewer's perception of the scene changes. The edits target the most important person's apparent emotion, character, status, or the scene's narrative, rather than swapping faces. On this dataset, the best zero-shot detector reaches 66.87% class-wise accuracy and human observers achieve 61.67%, so the manipulations largely evade both. The authors argue this exposes a gap: detectors trained on face swaps and object inpainting are not ready for semantic, person-centric manipulation.

What carries the argument

The key machinery is a two-stage VLM pipeline: first, a vision-language model (Gemini-2.0-Flash or ChatGPT-4o-latest) reads the image, identifies the most important person, and proposes six minimal edits, each expressed as a referring expression plus an edit instruction; second, an image-editing VLM (GPT-Image-1, Gemini-2.0-Flash-Image-Generation, or ICEdit) executes the edit while leaving everything else unchanged. The dataset is built on 86,952 real images from EMOTIC, PISC, PIPA, and PIC 2.0, and each manipulation is subsequently analyzed by a VLM for its perceptual and ethical impact.

What would settle it

A human-annotation study in which, for example, 200 edited images are given pixel-level masks by multiple annotators; if the dataset's auto-masks have low intersection over union with the human masks, the reported localization numbers are not measuring what they claim.

Watch

Extended reading notes

Core claim

The central claim is that current state-of-the-art deepfake detection methods and human observers cannot reliably detect person-centric, semantically meaningful image manipulations generated by vision-language models. The paper supports this by constructing MultiFakeVerse, in which each real image is paired with several manipulated versions produced from VLM-generated edit instructions targeting perception of the most important person. Benchmarking shows that AntiFakePrompt, the best zero-shot detector, achieves 66.87% class-wise accuracy and 55.55% F1 score, while humans reach only 61.67% accuracy with a 24.96% intersection over union on manipulation-level identification. After finetuning, CNNSpot and SIDA improve but still fall short of reliable detection.

Load-bearing premise

The paper's localization results rest on masks produced by thresholding pixel differences between each real and edited image, with no human validation that those masks actually cover the manipulated regions.

Editorial extensions

If this is right

  • Detectors trained on existing GAN- and inpainting-based datasets will systematically miss VLM-driven semantic edits, so new training or domain adaptation is needed.
  • The dataset provides a benchmark for both detection and localization of person-centric manipulations, with localization masks computed from pixel differences.
  • Human performance near chance suggests that person-centric semantic manipulations are a realistic threat not captured by current deepfake defenses.
  • Finetuning on MultiFakeVerse improves detection but does not bring it to reliable levels, indicating the task is not solved by simple supervised adaptation.

Reading between the lines

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

  • If the auto-generated localization masks are unreliable, the reported localization numbers may not reflect true localization quality; a human-annotation validation study could settle this.
  • The VLM-driven manipulation pipeline could be extended to video or audio-visual scenarios, where narrative-shifting edits may be even harder to detect.
  • The dataset could serve as a stress test for detector robustness to distribution shift, since the manipulations are semantically coherent rather than artifact-driven.
  • The low human accuracy on a 50-image sample hints that the broader population may struggle similarly, raising questions about how to communicate the existence of such edits.
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

3 major / 5 minor

Summary. The paper introduces MultiFakeVerse, a large-scale person-centric deepfake dataset of 845,286 images (87,245 real, 758,041 fake) created by using VLMs to propose and execute subtle semantic edits targeting emotions, narrative, and human-object interactions. The authors benchmark zero-shot and finetuned deepfake detectors (CNNSpot, TruFor, AntiFakePrompt, SIDA-13B), report a human study (61.67% accuracy), and evaluate forgery localization on SIDA-13B using masks automatically derived from pixel differences. The central claims are that the dataset is large, semantically meaningful, and hard for both detectors and humans.

Significance. If the dataset and its annotations are valid, MultiFakeVerse would be a valuable resource: it is substantially larger than most prior person-centric deepfake benchmarks, uses a novel VLM-driven manipulation pipeline, and provides initial evidence that current detectors and humans struggle on these subtle semantic edits. The paper also ships code and dataset links, and the zero-shot detection results (best 66.87% accuracy) are concrete falsifiable claims. However, the spatial-localization contribution currently rests on unvalidated, insufficiently documented mask generation, and the perceptual-impact analysis is self-referential; both issues need to be resolved before the dataset can serve as a reliable benchmark for localization or for claims about viewer perception.

major comments (3)
  1. [§3.2.1, §4.0.3, Conclusion] The forgery localization ground truth is not adequately supported. Section 3.2.1 states that masks are obtained by taking the mean squared pixel difference, thresholding to remove noise, and running connected-component analysis, but the threshold value, noise-filter parameters, and any alignment/resizing step are never reported. For GPT-Image-1 edits, Section 3.1.2 notes that the model has fixed output aspect ratios (1024×1024, 1536×1024, 1024×1536), and that it 'tends to edit in a few cases, untargeted regions'; pixel-wise differencing is only meaningful if the real and edited images are geometrically aligned, and no such preprocessing is described. Since the localization benchmark in Table 4 (SIDA-13B IoU 13.10/24.74, F1 19.92/39.40, AUC 14.06/37.53) and the conclusion's claim of 'the largest image-based dataset for spatial deepfake localization' depend entirely on these masks, the authors must either release the exact mask-generation code and parameters, validate the masks on a human-annotated subset, handle GPT-image geometry explicitly, or explicitly re-scope the dataset as detection-only without localization ground truth. As written, the localization results cannot be separated from ground-truth quality.
  2. [§3.2.4, Figure 3, Prompt 3.2] The perceptual-impact analysis is self-referential and should be disclosed as such. The edits are generated by Gemini-2.0-Flash-Image-Generation, and Prompt 3.2 asks Gemini-2.0-Flash to judge the resulting changes in emotion, identity, narrative, intent, and ethics. The word clouds in Figure 3 and the ethical-impact percentages (81% mild, 14.2% moderate, 0.3% severe) are therefore the editing model family's self-assessments, not independent measurements of human perception. The paper should reframe these as 'VLM-assessed' properties, add an explicit limitation noting the circularity, and ideally validate a subset with human raters before using the word clouds to claim that the manipulations are 'meaningful' or have particular ethical implications.
  3. [§4.0.2, Table 4] The finetuning results are reported inconsistently. The text states that 'we observe a performance improvement in both CNNSpot and SIDA-13B' and that CNNSpot surpasses SIDA-13B 'by 1.92%' in accuracy and 'by 1.97%' in F1-score. However, Table 4 shows that after finetuning, CNNSpot's overall accuracy drops from 50.02 to 45.88 and SIDA-13B's overall accuracy drops from 55.97 to 38.01; the F1 scores increase, but the accuracy gaps are 7.87 and 33.12 percentage points, respectively, not 1.92 and 1.97. The arrow for CNNSpot overall accuracy is also marked '↑' when the value decreases. The authors should clarify which metric the percentage differences refer to, correct the arrow and text, or revise the table so that the finetuning narrative matches the reported numbers.
minor comments (5)
  1. [§3.2.4] The text refers to 'Prompt 3.2.3' but the prompt is numbered 3.2; the cross-reference should be corrected.
  2. [§3.1.2, §3.2.7] The paper is ambiguous about which editing VLMs (Gemini, GPT-Image-1, ICEdit) contribute images to the final dataset of 758,041 fakes. Section 3.1.2 says Gemini emerged as the best after observing 22K generated images, yet Section 3.2.7 reports costs for GPT-Image-1 and ICEdit. Please clarify the per-model composition of the released dataset, because the localization-mask validity (see major comment above) depends on knowing whether GPT-Image-1 images with altered aspect ratios are included.
  3. [§3.2.6] The user study uses only 18 participants and 50 images; the authors should state that this is a pilot-level evaluation and consider reporting participant demographics or inter-rater agreement, which would help readers calibrate the 61.67% human accuracy figure.
  4. [Figure 3a] The x-axis label for the histogram of edited-area ratios is missing; it should be something like 'ratio of edited area to total image area' to match the text.
  5. [Table 4] The method name 'CnnSpot' is spelled inconsistently; elsewhere it is 'CNNSpot'. Please standardize the spelling.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the dataset benchmarks rest on independent human and detector evaluations; noted self-referential VLM analyses and mask-validity concerns are reliability issues, not circular reductions.

full rationale

The paper's central claims are empirical: (i) a large-scale dataset exists (845,286 images), (ii) state-of-the-art detectors and humans struggle on it (AntiFakePrompt at 66.87% class-wise accuracy; humans at 61.67%). These are supported by external benchmark models and a user study, not by the paper's own generative pipeline. The localization masks in §3.2.1 are auto-thresholded pixel-difference maps with unreported parameters and no human validation, and GPT-Image-1's fixed output aspect ratios in §3.1.2 may break pixel-wise differencing; this is a serious validity and reproducibility concern for the §4.0.3 localization benchmark and the 'largest localization dataset' claim, but it is not a circular derivation because SIDA-13B's predictions are independent of the mask-generation algorithm, so the IoU/F1/AUC numbers do not reduce to the input by construction. Similarly, the perceptual-impact analysis in §3.2.4 uses Gemini-2.0-Flash, the same model family that suggested and executed the edits; this is a self-evaluation weakness and the word-cloud results should be treated as descriptive rather than as predictions, but the paper does not use them as a formal derived result. No load-bearing self-citation or uniqueness-imported-from-authors pattern appears; prior-work citations to the authors' datasets in related work are contextual only. Overall, no step in the claimed derivation chain is equivalent to its inputs by construction.

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

The paper introduces no new physical or mathematical entities; its novel artifacts are data (images and computed masks) and a prompt taxonomy. The computed manipulation masks are listed in free_parameters because they function as unvalidated, hand-thresholded ground truth rather than as an independently evidenced entity. The three free parameters and four domain assumptions above cover the hand-chosen numbers and the load-bearing premises that the dataset's meaning, quality, and localization claims rest on.

free parameters (3)
  • Pixel-difference threshold for edit-mask extraction = not reported
    §3.2.1: masks used as localization ground truth are derived by thresholding mean squared pixel differences and connected-component analysis; the threshold value, noise cutoff, and post-processing are unspecified and the masks have no human validation.
  • Six target effects (naive, nonchalant, proud, remorseful, inexperienced, factual change) = 6 fixed categories
    Prompt 3.1: the dataset's definition of 'perceptual manipulation' is limited to six hand-chosen VLM targets; this choice shapes the entire dataset and is not justified by any perceptual study.
  • Selection of Gemini as primary editing VLM = Gemini-2.0-Flash-Image-Generation
    §3.1.2: Gemini was chosen after the authors visually inspected 22K generated images; GPT-Image-1 was excluded for editing untargeted regions and ICEdit for artifacts. This is a hand selection based on informal inspection, not a quantified metric.
assumptions (4)
  • domain assumption VLM-suggested edits are valid operationalizations of 'appears naive/proud/remorseful/inexperienced/nonchalant' and of 'factual change' as perceptual manipulations.
    Prompt 3.1: the six edit suggestions come from a VLM with no human verification that the suggested edit achieves the named perceptual effect; the dataset's motivation ('influence human perception of importance, intent, or narrative') inherits this unverified premise.
  • domain assumption Editing VLMs (especially Gemini) make minimal edits and leave untargeted regions unchanged.
    §3.1.2: the authors concede GPT-Image-1 edits untargeted regions and ICEdit produces artifacts; Gemini is claimed to be coherent 'with no changes to the untargeted regions' based only on visual inspection of a subset, not quantitative measurement.
  • domain assumption Pixel-difference connected components are an unbiased proxy for the semantically manipulated regions.
    §3.2.1: masks are computed rather than annotated; without human validation the localization benchmark in §4.0.3 may measure agreement with a pixel-difference heuristic rather than with perception of the edit.
  • ad hoc to paper The Gemini-2.0-Flash VLM is a reliable judge of the perceptual and ethical impact of the manipulations it helped create.
    Prompt 3.2 and §3.2.4: the same model family that suggested and executed the edits also scores emotion, identity, narrative, intent, and ethical impact; word-cloud claims such as 'manipulations are able to influence the individual's identity' are self-evaluations, not independent measurements.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Multiverse Through Deepfakes: The MultiFakeVerse Dataset of Person-Centric Visual and Conceptual Manipulations." pith.science (2026). https://pith.science/paper/AWWG6G2B

@misc{pith2026250600868,
  author       = {Pith},
  title        = {Pith review of: Multiverse Through Deepfakes: The MultiFakeVerse Dataset of Person-Centric Visual and Conceptual Manipulations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/AWWG6G2B}},
  note         = {Machine review of arXiv:2506.00868}
}
read the original abstract

The rapid advancement of GenAI technology over the past few years has significantly contributed towards highly realistic deepfake content generation. Despite ongoing efforts, the research community still lacks a large-scale and reasoning capability driven deepfake benchmark dataset specifically tailored for person-centric object, context and scene manipulations. In this paper, we address this gap by introducing MultiFakeVerse, a large scale person-centric deepfake dataset, comprising 845,286 images generated through manipulation suggestions and image manipulations both derived from vision-language models (VLM). The VLM instructions were specifically targeted towards modifications to individuals or contextual elements of a scene that influence human perception of importance, intent, or narrative. This VLM-driven approach enables semantic, context-aware alterations such as modifying actions, scenes, and human-object interactions rather than synthetic or low-level identity swaps and region-specific edits that are common in existing datasets. Our experiments reveal that current state-of-the-art deepfake detection models and human observers struggle to detect these subtle yet meaningful manipulations. The code and dataset are available on \href{https://github.com/Parul-Gupta/MultiFakeVerse}{GitHub}.

Figures

Figures reproduced from arXiv: 2506.00868 by the authors.

Figure 1
Figure 1. MultiFakeVerse. A brief overview of the proposed dataset. Here, we introduce subtle and profound person-centric deepfakes [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. (a) Depicts the MultiFakeVerse dataset generation pipeline. For details please see Section [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. The visualizations illustrate the characteristics and [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Analyzing the perceptual impact of manipulations in images. The edited regions are highlighted by yellow boxes. The analysis [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

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

  1. Explainable Deepfake Detection Challenge

    cs.CV 2026-07 conditional novelty 5.0 of 10

    A joint deepfake detection-and-explanation benchmark shows that classification accuracy and grounded, audience-appropriate explanations are separate capabilities, with Pixel Sleuth winning at 0.7612.

Reference graph

Works this paper leans on

49 extracted references · 23 canonical work pages · cited by 1 Pith paper

  1. [1]

    Rohan Anil and et al. 2023. Gemini: A Family of Highly Capable Multimodal Models.arXiv preprint arXiv:2312.11805(2023). https://arxiv.org/abs/2312. 11805

    arXiv 2023

  2. [2]

    Jordan J Bird and Ahmad Lotfi. 2024. Cifake: Image classification and explainable identification of ai-generated synthetic images.IEEE Access12 (2024), 15642– 15650

    2024

  3. [3]

    Tim Brooks, Aleksander Holynski, and Alexei A Efros. 2022. InstructPix2Pix: Learning to Follow Image Editing Instructions.arXiv preprint arXiv:2211.09800 (2022)

    arXiv 2022

  4. [4]

    Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, Sandhini Agarwal, Ariel Herbert-V oss, Gretchen Krueger, Tom Henighan, Rewon Child, Aditya Ramesh, Daniel Ziegler, Jeffrey Wu, Clemens Winter, Chris Hesse, Mark Chen, Eric Sigler, Mateusz Litwin, Scott G...

    work page 2020

  5. [5]

    Zhixi Cai, Shreya Ghosh, Aman Pankaj Adatia, Munawar Hayat, Abhinav Dhall, Tom Gedeon, and Kalin Stefanov. 2024. A V-Deepfake1M: A Large- Scale LLM-Driven Audio-Visual Deepfake Dataset. InProceedings of the 32nd ACM International Conference on Multimedia(Melbourne VIC, Australia)(MM ’24). Association for Computing Machinery, New York, NY , USA, 7414–7423....

    arXiv 2024

  6. [6]

    Zhixi Cai, Shreya Ghosh, Abhinav Dhall, Tom Gedeon, Kalin Stefanov, and Munawar Hayat. 2023. Glitch in the matrix: A large scale benchmark for content driven audio–visual forgery detection and localization.Computer Vision and Image Understanding236 (2023), 103818

    work page 2023

  7. [7]

    Zhixi Cai, Kalin Stefanov, Abhinav Dhall, and Munawar Hayat. 2022. Do You Really Mean That? Content Driven Audio-Visual Deepfake Dataset and Multi- modal Method for Temporal Forgery Localization. In2022 International Confer- ence on Digital Image Computing: Techniques and Applications (DICTA). 1–10. doi:10.1109/DICTA56598.2022.10034605

    arXiv 2022

  8. [8]

    Edresson Casanova, Christopher Shulby, Eren Gölge, Nicolas Michael Müller, Frederico Santos De Oliveira, Arnaldo Candido Jr., Anderson Da Silva Soares, Sandra Maria Aluisio, and Moacir Antonelli Ponti. 2021. SC-GlowTTS: An Efficient Zero-Shot Multi-Speaker Text-To-Speech Model. InInterspeech 2021. ISCA, 3645–3649

    work page 2021

Show all 49 references
  1. [9]

    You-Ming Chang, Chen Yeh, Wei-Chen Chiu, and Ning Yu. 2024. Antifake- Prompt: Prompt-Tuned Vision-Language Models are Fake Image Detectors. arXiv:2310.17419 [cs.CV] https://arxiv.org/abs/2310.17419

    2024 arXiv

  2. [10]

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

    2024

  3. [11]

    Bermano, Gal Chechik, and Daniel Cohen-Or

    Rinon Gal, Or Patashnik, Haggai Maron, Amit H. Bermano, Gal Chechik, and Daniel Cohen-Or. 2022. StyleGAN-NADA: CLIP-guided domain adaptation of image generators.ACM Trans. Graph.41, 4, Article 141 (July 2022), 13 pages. doi:10.1145/3528223.3530164

    2022

  4. [12]

    Songwei Ge, Thomas Hayes, Harry Yang, Xi Yin, Guan Pang, David Jacobs, Jia-Bin Huang, and Devi Parikh. 2022. Long video generation with time-agnostic vqgan and time-sensitive transformer. InEuropean Conference on Computer Vision. Springer, 102–118

    2022

  5. [13]

    Fabrizio Guillaro, Davide Cozzolino, Avneesh Sud, Nicholas Dufour, and Luisa Verdoliva. 2023. TruFor: Leveraging All-Round Clues for Trustworthy Image Forgery Detection and Localization. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)...

    2023

  6. [14]

    Martin Heusel, Hubert Ramsauer, Thomas Unterthiner, Bernhard Nessler, and Sepp Hochreiter. 2017. GANs Trained by a Two Time-Scale Update Rule Converge to a Local Nash Equilibrium. InAdvances in Neural Informa- tion Processing Systems, V ol. 30. https://papers.nips.cc/paper/201...

    2017

  7. [15]

    Zhenglin Huang, Jinwei Hu, Xiangtai Li, Yiwei He, Xingyu Zhao, Bei Peng, Baoyuan Wu, Xiaowei Huang, and Guangliang Cheng. 2024. SIDA: Social Media Image Deepfake Detection, Localization and Explanation with Large Multimodal Model.arXiv preprint arXiv:2412.04292(2024)

    2024 arXiv

  8. [16]

    Tero Karras, Samuli Laine, and Timo Aila. 2019. A style-based generator ar- chitecture for generative adversarial networks. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition. 4401–4410

    2019

  9. [17]

    Ronak Kosti, Jose M Alvarez, Adria Recasens, and Agata Lapedriza. 2019. Con- text based emotion recognition using emotic dataset.IEEE transactions on pattern analysis and machine intelligence42, 11 (2019), 2755–2766

    2019

  10. [18]

    Junnan Li, Yongkang Wong, Qi Zhao, and Mohan S Kankanhalli. 2017. Dual- glance model for deciphering social relationships. InProceedings of the IEEE international conference on computer vision. 2650–2659

    2017

  11. [19]

    Si Liu, Zitian Wang, Yulu Gao, Lejian Ren, Yue Liao, Guanghui Ren, Bo Li, and Shuicheng Yan. 2022. Human-Centric Relation Segmentation: Dataset and Solution.IEEE Transactions on Pattern Analysis and Machine Intelligence44, 9 (2022), 4987–5001. doi:10.1109/TPAMI.2021.3075846

    2022

  12. [20]

    Brandon B May, Kirill Trapeznikov, Shengbang Fang, and Matthew Stamm. 2023. Comprehensive dataset of synthetic and manipulated overhead imagery for de- velopment and evaluation of forensic tools. InProceedings of the 2023 ACM Workshop on Information Hiding and Multimedia Secur...

    2023

  13. [21]

    Changtao Miao, Qi Chu, Zhentao Tan, Zhenchao Jin, Tao Gong, Wanyi Zhuang, Yue Wu, Bin Liu, Honggang Hu, and Nenghai Yu. 2023. Multi-spectral Class Center Network for Face Manipulation Detection and Localization.arXiv preprint arXiv:2305.10794(2023). https://arxiv.org/abs/2305.10794

    2023 arXiv

  14. [22]

    Kartik Narayan, Harsh Agarwal, Kartik Thakral, Surbhi Mittal, Mayank Vatsa, and Richa Singh. 2023. DF-Platter: Multi-Face Heterogeneous Deepfake Dataset. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). 9739–9748

    2023

  15. [23]

    Thanh Thi Nguyen, Quoc Viet Hung Nguyen, Dung Tien Nguyen, Duc Thanh Nguyen, Thien Huynh-The, Saeid Nahavandi, Thanh Tam Nguyen, Quoc-Viet Pham, and Cuong M. Nguyen. 2019. Deep Learning for Deepfakes Creation and Detection: A Survey.arXiv preprint arXiv:1909.11573(2019). https...

    2019 arXiv

  16. [24]

    Adam Novozamsky, Babak Mahdian, and Stanislav Saic. 2020. IMD2020: A large- scale annotated dataset tailored for detecting manipulated images. InProceedings of the IEEE/CVF winter conference on applications of computer vision workshops. 71–80

    2020

  17. [25]

    OpenAI. 2024. GPT-4o System Card. https://arxiv.org/abs/2410.21276. Accessed: 2025-05-31

    2024 arXiv

  18. [26]

    Anisha Pal, Julia Kruk, Mansi Phute, Manognya Bhattaram, Diyi Yang, Duen Horng Chau, and Judy Hoffman. 2024. Semi-Truths: A Large-Scale Dataset of AI-Augmented Images for Evaluating Robustness of AI-Generated Image detectors.Advances in Neural Information Processing Systems37 ...

    2024

  19. [27]

    Viraj Prabhu, Sriram Yenamandra, Prithvijit Chattopadhyay, and Judy Hoffman

  20. [28]

    Md Awsafur Rahman, Bishmoy Paul, Najibul Haque Sarker, Zaber Ibn Abdul Hakim, and Shaikh Anowarul Fattah. 2023. Artifact: A large-scale dataset with artificial and factual images for generalizable and robust synthetic image detection. In2023 IEEE International Conference on Im...

    2023

  21. [29]

    Vishal Kumar Sharma, Rakesh Garg, and Quentin Caudron. 2024. A system- atic literature review on deepfake detection techniques.Multimedia Tools and Applications(2024). doi:10.1007/s11042-024-19906-1

    2024 doi

  22. [30]

    Kai Shen, Zeqian Ju, Xu Tan, Yanqing Liu, Yichong Leng, Lei He, Tao Qin, Sheng Zhao, and Jiang Bian. 2023. Naturalspeech 2: Latent diffusion models are natural and zero-shot speech and singing synthesizers.arXiv preprint arXiv:2304.09116 (2023)

    2023 arXiv

  23. [31]

    Joel Stehouwer, Hao Dang, Feng Liu, Xiaoming Liu, and Anil Jain. 2019. On the detection of digital face manipulation.arXiv(2019), arXiv–1910

    2019

  24. [32]

    Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timothée Lacroix, Baptiste Rozière, Naman Goyal, Eric Hambro, Faisal Azhar, et al. 2023. Llama: Open and efficient foundation language models.arXiv preprint arXiv:2302.13971(2023)

    2023 arXiv

  25. [33]

    Hugo Touvron, Louis Martin, Kevin Stone, Peter Albert, Amjad Almahairi, Yas- mine Babaei, Nikolay Bashlykov, Soumya Batra, Prajjwal Bhargava, Shruti Bhos- ale, Dan Bikel, Lukas Blecher, Cristian Canton Ferrer, Moya Chen, Guillem Cucurull, David Esiobu, Jude Fernandes, Jeremy F...

    2023 arXiv

  26. [34]

    Soroush V osoughi, Deb Roy, and Sinan Aral. 2018. The Spread of True and False News Online.Science359, 6380 (2018), 1146–1151. https://news.mit.edu/2018/ study-twitter-false-news-travels-faster-true-stories-0308 Accessed: 2024-05-30

    2018

  27. [35]

    Jia Wang, Jie Hu, Xiaoqi Ma, Hanghang Ma, Xiaoming Wei, and Enhua Wu. 2025. Image Editing with Diffusion Models: A Survey.arXiv preprint arXiv:2504.13226 (2025). https://arxiv.org/abs/2504.13226

    2025 arXiv

  28. [36]

    Run Wang, Felix Juefei-Xu, Lei Ma, Xiaofei Xie, Yihao Huang, Jian Wang, and Yang Liu. 2019. Fakespotter: A simple yet robust baseline for spotting ai-synthesized fake faces.arXiv preprint arXiv:1909.06122(2019). Conference, June 2025, Washington, DC, USA Gupta et al

    2019 arXiv

  29. [37]

    Sheng-Yu Wang, Oliver Wang, Richard Zhang, Andrew Owens, and Alexei A Efros. 2020. CNN-generated images are surprisingly easy to spot... for now. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition. 8695–8704

    2020

  30. [38]

    Bovik, H.R

    Zhou Wang, A.C. Bovik, H.R. Sheikh, and E.P. Simoncelli. 2004. Image quality assessment: from error visibility to structural similarity.IEEE Transactions on Image Processing13, 4 (April 2004), 600–612

    2004

  31. [39]

    Zhikan Wang, Zhongyao Cheng, Jiajie Xiong, Xun Xu, Tianrui Li, Bharadwaj Veeravalli, and Xulei Yang. 2024. A Timely Survey on Vision Transformer for Deepfake Detection.arXiv preprint arXiv:2405.08463(2024). https://arxiv.org/ abs/2405.08463

    2024 arXiv

  32. [40]

    Claire Wardle. 2019. The Disturbing World of Deepfake Pornography.WIRED (October 2019). https://www.wired.com/story/deepfakes-pornography Accessed: 2024-05-30

    2019

  33. [41]

    Shaoan Xie, Zhifei Zhang, Zhe Lin, Tobias Hinz, and Kun Zhang. 2023. Smart- Brush: Text and Shape Guided Object Inpainting with Diffusion Model. InPro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recogni- tion (CVPR). 22428–22437. doi:10.1109/CVPR52729.2...

    2023

  34. [42]

    Beichen Zhang, Pan Zhang, Xiaoyi Dong, Yuhang Zang, and Jiaqi Wang. 2024. Long-CLIP: Unlocking the Long-Text Capability of CLIP.arXiv preprint arXiv:2403.15378(2024)

    2024 arXiv

  35. [43]

    Ning Zhang, Manohar Paluri, Yaniv Taigman, Rob Fergus, and Lubomir Bourdev

  36. [44]

    Zechuan Zhang, Ji Xie, Yu Lu, Zongxin Yang, and Yi Yang. 2025. In-Context Edit: Enabling Instructional Image Editing with In-Context Generation in Large Scale Diffusion Transformer. arXiv:2504.20690 [cs.CV] https://arxiv.org/abs/ 2504.20690

    2025 arXiv

  37. [45]

    Nan Zhong, Yiran Xu, Zhenxing Qian, and Xinpeng Zhang. 2023. Rich and poor texture contrast: A simple yet effective approach for ai-generated image detection. CoRR(2023)

    2023

  38. [46]

    Mingjian Zhu, Hanting Chen, Qiangyu Yan, Xudong Huang, Guanyu Lin, Wei Li, Zhijun Tu, Hailin Hu, Jie Hu, and Yunhe Wang. 2023. Genimage: A million-scale benchmark for detecting ai-generated image.Advances in Neural Information Processing Systems36 (2023), 77771–77782

    2023

  39. [47]

    Giada Zingarini, Davide Cozzolino, Riccardo Corvi, Giovanni Poggi, and Luisa Verdoliva. 2024. M3Dsynth: A dataset of medical 3D images with AI-generated local manipulations. InICASSP 2024-2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). I...

    2024

  40. [2015]

    arXiv:1501.05703 [cs.CV] https://arxiv.org/abs/1501.05703

    Beyond Frontal Faces: Improving Person Recognition Using Multiple Cues. arXiv:1501.05703 [cs.CV] https://arxiv.org/abs/1501.05703

    arXiv

  41. [2023]

    InNeural Information Processing Systems (NeurIPS)

    LANCE: Stress-testing Visual Models by Generating Language-guided Counterfactual Images. InNeural Information Processing Systems (NeurIPS)

Pith tools

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