Pith. sign in

REVIEW 2 major objections 4 minor 1 cited by

Re-Processing Images Exposes AI Fakes Across All Generator Families

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · glm-5.2

2026-07-10 03:11 UTC pith:JGGG3ZIX

load-bearing objection The headline comparison is confounded by backbone choice: GenRes++ uses PE-Core-G14-448 while all baselines use CLIP, and the paper never re-runs baselines with PE-Core to isolate the generative residual contribution from the backbone upgrade. the 2 major comments →

arxiv 2607.08674 v1 pith:JGGG3ZIX submitted 2026-07-09 cs.CV

Do Transformations Reveal the Truth? Generative Residual Learning for Generalized AI-Generated Image Detection

Kutub Uddin , Nusrat Tasnim , Awais Khan , Mohammad Umar Farooq , Khalid Malik This is my paper
classification cs.CV
keywords generativedetectiongenresmethodsmultipleai-generatedaigifeatures
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper argues that AI-generated images carry a hidden signature that is invisible when you look at the image alone but becomes detectable when you run the image through a second generative model — a super-resolver, denoiser, or face restorer. Real photos and synthetic images respond differently to this secondary processing because the synthetic image already encodes the statistical biases of its source generator, and those biases interact with the assumptions of the transform in a measurable way. The authors call this differential response a generative residual. They build a detector around it: encode the original and its transformed variants with a frozen vision transformer, model the multiplicative cross-feature interactions between original and transformed embeddings with a neural tensor network, and use cross-attention to aggregate signals from five complementary transforms. The resulting system detects images from 19 generator families it never saw during training, achieving 95.7% accuracy and 99.1% average precision, surpassing all prior methods on the standard cross-generator benchmark.

Core claim

The central discovery is that the generative residual — the discrepancy between an image and its output under a secondary generative transform — is a cross-generator detection signal. Unlike generator-specific spectral or upsampling artifacts, this differential response arises from the interaction between two generative priors (the source generator and the transform) and is empirically present across both GAN and diffusion families. By explicitly modeling multiplicative cross-feature dependencies between original and transformed embeddings via a neural tensor network, and aggregating across five diverse transforms with attention, the detector generalizes to unseen generators without learning

What carries the argument

The architecture has four components: (1) a frozen PE-Core vision encoder with LoRA rank-6 adaptation producing 1536-dim embeddings projected to 256 dims; (2) five fixed generative transforms — EnlightenGAN, GFPGAN, Real-ESRGAN, FFDNet (σ=25), and CodeFormer — applied to every input; (3) a cross-attention aggregation module treating the original embedding as query and the five transformed embeddings as keys/values; (4) a neural tensor network computing f = tanh(z₀ᵀ W g_agg + [z₀; g_agg]ᵀ V + b) with W ∈ ℝ^{256×256×128}, producing a 128-dim relational vector fed to a binary classification head. Training uses only ProGAN images; the LoRA adapters, projection, CCA, NTN, and head are the only训练d

Load-bearing premise

The paper assumes that the differential response of synthetic images under secondary generative processing is a universal property that will hold for any future generator architecture, not just the 19 in the benchmark. If a new generator produces images whose statistical structure is indistinguishable from natural images under all five transforms, the detection signal vanishes.

What would settle it

Construct or discover a generator whose images produce transformed outputs statistically indistinguishable from transformed real images across all five transforms (EnlightenGAN, GFPGAN, Real-ESRGAN, FFDNet, CodeFormer). If the generative residual between original and all five transformed variants is below the detection threshold for such images, the core mechanism fails and the framework cannot detect that generator.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • If generative residuals are truly a universal property of synthetic images, then any future generator — regardless of architecture — should be detectable without retraining, as long as the transform set exposes the residual. This would shift the detection arms race from matching generators to selecting transforms.
  • The cross-attention aggregation mechanism implies that not all transforms are equally informative for every image. An adaptive transform-selection policy could reduce the five-transform computational overhead while preserving detection power.
  • If the residual signal generalizes to video, temporal consistency of generative residuals across frames could provide a stronger and more robust detection cue than per-frame analysis, since temporal artifacts would compound.
  • The framework's reliance on fixed, manually selected transforms creates a known attack surface: a generator designed to minimize residual discrepancy under these specific five transforms could evade detection, motivating adversarial or learned transform sets.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 4 minor

Summary. This manuscript proposes GenRes and GenRes++, frameworks for detecting AI-generated images (AIGI) by modeling the differential response of real versus synthetic images to secondary generative transformations (e.g., denoising, super-resolution). The core idea is that applying a second generative model to an AIGI produces characteristic residual discrepancies due to the interaction between the source generator's biases and the transformation network's priors. GenRes models these residuals using a Neural Tensor Network (NTN) on a single transform, while GenRes++ extends this by using a Cross-Attention Aggregation (CCA) module to pool features from N=5 complementary transforms. Both models use a frozen PE-Core vision encoder with LoRA fine-tuning. Evaluated on the UniversalFakeDetect benchmark (trained on ProGAN, tested on 19 unseen generators), GenRes++ achieves 95.7% mACC and 99.1% mAP, outperforming existing baselines. The experimental protocol is standard, and the ablation studies generally support the architectural design choices.

Significance. The problem of cross-generator AIGI detection is practically consequential and actively studied. The paper's central premise—that the differential response of images to secondary generative processing provides a generator-agnostic signal—is intuitive and well-motivated. The empirical results are strong, and the ablations (Tables 4-7, Figs. 4-6) provide falsifiable evidence for the contribution of individual transforms, LoRA rank, and fusion strategies. However, the significance of the proposed generative residual framework over a simple backbone upgrade is currently obscured by a major confounding variable in the experimental design, which must be addressed before the central claim can be fully accepted.

major comments (2)
  1. The headline comparison in Tables 1 and 2 is confounded by the choice of backbone. GenRes++ uses PE-Core-G14-448, while all baselines (UniFD, C2P-CLIP, FreLens, RINE, etc.) use CLIP-based backbones. Table 5 shows that swapping CLIP ViT-L/14 for PE-Core under identical GenRes++ training yields a 4.4% ACC gain (91.3% to 95.7%). Furthermore, Table 4 shows that PE-Core with zero transforms already reaches 91.4% ACC / 96.1% AP, which is competitive with several baselines despite using no relational residual modeling. The critical missing experiment is: what happens if a strong baseline (e.g., FreLens or C2P-CLIP) is re-run with PE-Core as its backbone? If a baseline gains even 1-2% from PE-Core, it would match or exceed GenRes++, collapsing the claim that the proposed framework outperforms existing methods. The transforms do add value over PE-Core alone (91.4% to 95.7%), but this is a within-
  2. method ablation, not a cross-method comparison. The paper never establishes that 'PE-Core + generative residuals' beats 'PE-Core + best existing method.' Without this control, the improvement over baselines could be substantially or entirely attributable to the backbone rather than the proposed framework. The authors must isolate the generative residual contribution from the backbone upgrade to validate the central claim.
minor comments (4)
  1. §3.2.1: The text states 'd enc = 1536' for the PE-Core-G14-448 encoder, but later in §3.2.2 states 'Z = [z1, ..., zN] in R^{N x d} (d = 256)'. It would help to clarify explicitly that the bottleneck projection layer reduces the dimensionality from 1536 to 256 before the CCA module.
  2. Table 3: The inference time for GenRes++ is reported as 4625.6 ms per image. This is quite high for a detection task. While the authors acknowledge this limitation in §5, it would be beneficial to briefly discuss in §4.4 whether this latency is dominated by the sequential application of the five transforms or the multi-branch encoding, and whether the transforms could be parallelized in practice.
  3. §4.5.5 (Leave-One-Out): The text mentions 'removing FFDNet (ACC: 92.3%, AP: 96.4%), indicating that VQ-prior-based face restoration artifacts are highly discriminative.' However, FFDNet is a denoising network, not a VQ-prior-based face restoration model. CodeFormer is the VQ-prior-based model. This appears to be a mislabeling of the transforms in the analysis.
  4. Figure 3: The caption states 'Average PSNR/SSIM between the original and each transform... for the real and fake samples.' It is unclear if these averages are computed over the entire dataset or just the single real and fake sample shown in the figure. Clarifying this would improve reproducibility.

Circularity Check

0 steps flagged

No circularity: GenRes++ is evaluated against an external benchmark with unseen generators, uses external pre-trained transforms, and no prediction reduces to a fitted constant or self-citation chain.

full rationale

The paper's central claim — that GenRes++ achieves 95.7% mACC and 99.1% mAP across 19 unseen generative models — is evaluated against the external UniversalFakeDetect benchmark [32], with generators spanning GAN, diffusion, and other families that are entirely unseen during training (only ProGAN is used for training). The five generative transforms (EnlightenGAN, GFPGAN, Real-ESRGAN, FFDNet, CodeFormer) are external pre-trained models not developed by the authors. The NTN and CCA modules are standard architectures from [28] and [29] respectively. The training objective (Eq. 5) is standard binary cross-entropy. No equation in the derivation chain reduces to its own input by construction: the NTN output (Eq. 3) is a bilinear interaction between learned embeddings, not a definitional identity; the CCA aggregation (Eqs. 1-2) uses learned projection matrices; the LoRA adaptation follows the standard formulation W' = W + (α/r)AB. The authors do cite their own prior work [6,7,11,12,17,20-24] for context and motivation, but these citations are not load-bearing for the mathematical derivation or the empirical claims — the core premise (that real and synthetic images respond differently under secondary generative processing) is justified by citing independent work [18,19,43] and validated empirically on the external benchmark. The ablation studies (Tables 4-7, Figs. 4-6) are within-method comparisons that do not involve fitting a parameter to test data and then predicting the same data. The backbone comparison (Table 5) is a legitimate ablation, not a circular argument. While the skeptic correctly notes that the backbone choice (PE-Core vs. CLIP) confounds the cross-method comparison, this is a correctness/experimental design concern, not circularity — the paper does not define its output in terms of its input or rename a fitted quantity as a prediction. The derivation is self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

8 free parameters · 4 axioms · 0 invented entities

The paper introduces no new physical entities or postulated objects. The 'generative residual' is a named phenomenon, not a new entity. Free parameters are standard hyperparameters selected via ablation. The axioms are domain assumptions about image statistics, not mathematical postulates.

free parameters (8)
  • LoRA rank r = 6
    Selected empirically from {2,4,6,8,12,16} via ablation (Fig. 4).
  • LoRA alpha = 8
    Stated as fixed throughout; no ablation provided.
  • NTN output dimension k = 128
    Stated as fixed; no ablation provided.
  • Projection dimension d = 256
    Stated as fixed; no ablation provided.
  • Number of attention heads = 4
    Stated as fixed; no ablation provided.
  • Number of transforms N = 5
    Selected empirically from {0,...,8} via ablation (Fig. 6).
  • Learning rate = 4e-4
    Stated hyperparameter.
  • Weight decay = 0.05
    Stated hyperparameter.
axioms (4)
  • domain assumption Real and AI-generated images respond differently when processed by a second generative model.
    Central premise stated in §3.1; empirically supported by results but not theoretically derived.
  • domain assumption This differential response holds across generator families (GAN and diffusion).
    Stated in §3.1 as a 'general property'; validated on 19 generators but not proven in general.
  • domain assumption Generative residuals manifest as correlated perturbations across feature dimensions, requiring multiplicative (bilinear) rather than additive modeling.
    Stated in §3.1 to motivate NTN; supported by fusion-function ablation (Fig. 5) showing NTN outperforms subtraction/concatenation.
  • domain assumption PE-Core provides richer low-level texture representations than CLIP or DINO for this task.
    Invoked in §4.5.3 to explain backbone superiority; supported by Table 5 but not independently verified.

pith-pipeline@v1.1.0-glm · 21583 in / 2724 out tokens · 321202 ms · 2026-07-10T03:11:14.019839+00:00 · methodology

0 comments
read the original abstract

The rapid advancement of generative AI has enabled the creation of highly realistic deepfake media, posing significant threats, including misinformation, digital identity theft, fraud, and manipulation of public opinion. AI-generated image (AIGI) detection is reliably challenging due to the diversity of generative methods and the subtle artifacts they leave behind. In this work, we propose GenRes, a novel framework for generative residual learning via a neural tensor network, which models fine-grained relational features between original and transformed samples to enhance generalization. To address scenarios involving multiple generative transformations, we introduce GenRes++, which employs a learnable attention mechanism to aggregate relational features across multiple transformed samples and enables the model to focus on the most informative cues. Both models leverage PE-Core as a feature extractor, providing generalized and semantically rich embeddings that improve cross-domain performance and enable the detection of AIGI generated by unseen methods. Comprehensive experiments on multiple benchmark datasets demonstrate that the proposed GenRes++ approach outperforms existing methods.

Figures

Figures reproduced from arXiv: 2607.08674 by Awais Khan, Khalid Malik, Kutub Uddin, Mohammad Umar Farooq, Nusrat Tasnim.

Figure 1
Figure 1. Figure 1: Comparison between traditional AIGI detection and the [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Overview of the proposed GenRes + + framework. The original image I0 and its generative transformed variants {Ii} N i=1 are encoded using a shared PE-CORE encoder with LoRA adaptation, followed by a projection into a compact feature space. The transformed features are aggregated via a CCA module to produce a unified representation gagg. A NTN then models the bilinear relational interaction between the orig… view at source ↗
Figure 3
Figure 3. Figure 3: Qualitative comparison of five generative transforms applied to one real and one fake image (BigGAN). Average PSNR/SSIM [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Ablation on LoRA rank r. Bars show ACC (%) and the dashed line shows AP (%). The best configuration (r=6) is highlighted in green [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Radar chart comparing fusion functions in the NTN [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Effect of the number of generative transforms [PITH_FULL_IMAGE:figures/full_fig_p008_6.png] view at source ↗

discussion (0)

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

Forward citations

Cited by 1 Pith paper

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

  1. Uncertainty-Aware Deepfake Detection via Multi-View Structural Learning

    cs.CV 2026-07 conditional novelty 6.0

    A three-stream detector with disagreement-calibrated evidential uncertainty cuts calibration error roughly in half on four of five cross-dataset deepfake benchmarks.

Reference graph

Works this paper leans on

73 extracted references · 73 canonical work pages · cited by 1 Pith paper · 7 internal anchors

  1. [1]

    Progressive Growing of GANs for Improved Quality, Stability, and Variation

    Tero Karras, Timo Aila, Samuli Laine, and Jaakko Lehtinen. Progressive growing of gans for improved quality, stability, and variation.arXiv preprint arXiv:1710.10196, 2017. 1, 3, 5, 6

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  2. [2]

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

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising dif- fusion probabilistic models.Advances in neural information processing systems, 33:6840–6851, 2020. 1, 3

    work page 2020

  3. [3]

    Taming transformers for high-resolution image synthesis

    Patrick Esser, Robin Rombach, and Bj ¨orn Ommer. Taming transformers for high-resolution image synthesis. InCVPR,

    work page

  4. [4]

    AI Deepfakes Surge: $200 Million Lost, 2025

    eSecurityPlanet.com. AI Deepfakes Surge: $200 Million Lost, 2025. [Online; accessed 2025]. 1

    work page 2025

  5. [5]

    Deepfake-related fraud forecast to hit $40b by 2027,

    Eftsure. Deepfake-related fraud forecast to hit $40b by 2027,

    work page 2027

  6. [6]

    [Online; accessed 2025]. 1

    work page 2025

  7. [7]

    A com- prehensive survey, large-scale empirical study, and future in- sights on generalization, robustness, and explainability of ai- generated image detection.SSRN, 2026

    Nusrat Tasnim, Kutub Uddin, and Khalid Malik. A com- prehensive survey, large-scale empirical study, and future in- sights on generalization, robustness, and explainability of ai- generated image detection.SSRN, 2026. 1

    work page 2026

  8. [8]

    Adversarial attacks on audio deep- fake detection: A benchmark and comparative study.Pro- ceedings of the BMVC 2025 Workshop on Secure and Robust Biometrics Systems, 2025

    Kutub Uddin, Muhammad Umar Farooq, Awais Khan, and Khalid Mahmood Malik. Adversarial attacks on audio deep- fake detection: A benchmark and comparative study.Pro- ceedings of the BMVC 2025 Workshop on Secure and Robust Biometrics Systems, 2025. 1

    work page 2025

  9. [9]

    Leveraging frequency analysis for deep fake image recognition

    Joshua Frank, Thorsten Eisenhofer, Lea Sch ¨onherr, Andreas Fischer, Dorothea Kolossa, and Thorsten Holz. Leveraging frequency analysis for deep fake image recognition. InIn- ternational Conference on Machine Learning, pages 3247–

    work page

  10. [10]

    Watch your up-convolution: Cnn based generative deep neural net- works are failing to reproduce spectral distributions

    Ricard Durall, Margret Keuper, and Janis Keuper. Watch your up-convolution: Cnn based generative deep neural net- works are failing to reproduce spectral distributions. InPro- ceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 7890–7899, 2020. 1, 2

    work page 2020

  11. [11]

    Rethinking the up-sampling op- erations in cnn-based generative network for generalizable deepfake detection

    Chuangchuang Tan, Yao Zhao, Shikui Wei, Guanghua Gu, Ping Liu, and Yunchao Wei. Rethinking the up-sampling op- erations in cnn-based generative network for generalizable deepfake detection. InProceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, pages 28130–28139, 2024. 1, 2, 5, 6

    work page 2024

  12. [12]

    Diversity matters: Dataset diversification and dual-branch network for generalized ai-generated image detection

    Nusrat Tasnim, Kutub Uddin, and Khalid Malik. Diversity matters: Dataset diversification and dual-branch network for generalized ai-generated image detection. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 1420–1430, 2026. 1

    work page 2026

  13. [13]

    Sheild: A secure and highly enhanced integrated learning for robust deepfake detection against adversarial attacks

    Kutub Uddin, Awais Khan, Muhammad Umar Farooq, and Khalid Mahmood Malik. Sheild: A secure and highly enhanced integrated learning for robust deepfake detection against adversarial attacks. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 1502– 1511, 2025. 1

    work page 2025

  14. [14]

    Deep learning-based counter anti-forensic of gan-based attack in hevc compressed domain using coding pattern analysis.Ex- pert Systems with Applications, 233:120912, 2023

    Kutub Uddin, Yoonmo Yang, and Byung Tae Oh. Deep learning-based counter anti-forensic of gan-based attack in hevc compressed domain using coding pattern analysis.Ex- pert Systems with Applications, 233:120912, 2023. 1

    work page 2023

  15. [15]

    Towards uni- versal fake image detectors that generalize across genera- tive models

    Utkarsh Ojha, Yuheng Li, and Yong Jae Lee. Towards uni- versal fake image detectors that generalize across genera- tive models. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 24480– 24489, 2023. 1, 2, 3, 5, 6

    work page 2023

  16. [16]

    Raising the bar of ai-generated image detection with clip

    Davide Cozzolino, Giovanni Poggi, Riccardo Corvi, Matthias Nießner, and Luisa Verdoliva. Raising the bar of ai-generated image detection with clip. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 4356–4366, 2024. 1, 2

    work page 2024

  17. [17]

    C2p-clip: Inject- ing category common prompt in clip to enhance generaliza- tion in deepfake detection

    Chuangchuang Tan, Renshuai Tao, Huan Liu, Guanghua Gu, Baoyuan Wu, Yao Zhao, and Yunchao Wei. C2p-clip: Inject- ing category common prompt in clip to enhance generaliza- tion in deepfake detection. InProceedings of the AAAI Con- ference on Artificial Intelligence, volume 39, pages 7184– 7192, 2025. 1, 2, 3, 5, 6

    work page 2025

  18. [18]

    Face2parts: Exploring coarse-to-fine inter-regional facial de- pendencies for generalized deepfake detection.IEEE Access, 14:55111–55125, 2026

    Kutub Uddin, Nusrat Tasnim, and Byung Tae Oh. Face2parts: Exploring coarse-to-fine inter-regional facial de- pendencies for generalized deepfake detection.IEEE Access, 14:55111–55125, 2026. 1

    work page 2026

  19. [19]

    A robust open-set multi-instance learning for defending adversarial attacks in digital image.IEEE Trans- actions on Information Forensics and Security, 19:2098– 2111, 2023

    Kutub Uddin, Yoonmo Yang, Tae Hyun Jeong, and Byung Tae Oh. A robust open-set multi-instance learning for defending adversarial attacks in digital image.IEEE Trans- actions on Information Forensics and Security, 19:2098– 2111, 2023. 2

    work page 2098

  20. [20]

    Counter- act against gan-based attacks: A collaborative learning ap- proach for anti-forensic detection.Applied Soft Computing, 153:111287, 2024

    Kutub Uddin, Tae Hyun Jeong, and Byung Tae Oh. Counter- act against gan-based attacks: A collaborative learning ap- proach for anti-forensic detection.Applied Soft Computing, 153:111287, 2024. 2

    work page 2024

  21. [21]

    Nusrat Tasnim, Kutub Uddin, and Khalid Mahmood Malik. Ai-generated image detection: An empirical study and future research directions.Proceedings of the BMVC 2025 Work- shop on Media Authenticity in the Age of Artificial Intelli- gence, 2025. 2

    work page 2025

  22. [22]

    Grex-bench: Bench- marking generalization, robustness, and explainability in ai- generated image detection

    Nusrat Tasnim, Khalid Malik, et al. Grex-bench: Bench- marking generalization, robustness, and explainability in ai- generated image detection. 2026. 2

    work page 2026

  23. [23]

    Trace: Training-free partial audio deepfake detection via em- bedding trajectory analysis of speech foundation models

    Muhammad Umar Farooq, Kutub Uddin, Khalid Malik, et al. Trace: Training-free partial audio deepfake detection via em- bedding trajectory analysis of speech foundation models. In Proceedings of the IEEE/CVF Conference on Computer Vi- sion and Pattern Recognition, pages 7405–7414, 2026. 2

    work page 2026

  24. [24]

    Dual-Branch Gated Fusion for Open-Set Audio Deepfake Source Tracing

    Awais Khan, Kutub Uddin, and Khalid Malik. Dual-branch gated fusion for open-set audio deepfake source tracing. arXiv preprint arXiv:2606.10223, 2026. 2

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  25. [25]

    Advbench: A comprehensive benchmark of adver- sarial attacks on deepfake detectors in real-world consumer applications.Authorea Preprints, 2025

    Kutub Uddin, Muhammad Farooq, Awais Khan, Muhammad Saeed, Ijaz Ul Haq, Nusrat Tasnim, and Khalid Mahmood Malik. Advbench: A comprehensive benchmark of adver- sarial attacks on deepfake detectors in real-world consumer applications.Authorea Preprints, 2025. 2

    work page 2025

  26. [26]

    Adversarial perturbations fool deepfake detectors

    Apurva Gandhi and Shomik Jain. Adversarial perturbations fool deepfake detectors. InInternational joint conference on neural networks (IJCNN), pages 1–8. IEEE, 2020. 2

    work page 2020

  27. [27]

    Gan-generated image detection with self-attention mecha- nism against gan generator defect.IEEE Journal of Selected Topics in Signal Processing, 14(5):969–981, 2020

    Zhongjie Mi, Xinghao Jiang, Tanfeng Sun, and Ke Xu. Gan-generated image detection with self-attention mecha- nism against gan generator defect.IEEE Journal of Selected Topics in Signal Processing, 14(5):969–981, 2020. 2

    work page 2020

  28. [28]

    Improved denoising diffusion probabilistic models

    Alexander Quinn Nichol and Prafulla Dhariwal. Improved denoising diffusion probabilistic models. InInternational conference on machine learning, pages 8162–8171. PMLR,

    work page

  29. [29]

    Reasoning with neural tensor networks for knowledge base completion

    Richard Socher, Danqi Chen, Christopher D Manning, and Andrew Ng. Reasoning with neural tensor networks for knowledge base completion. InNeurIPS, 2013. 2, 3

    work page 2013

  30. [30]

    Atten- tion is all you need

    Ashish Vaswani, Noam Shazeer, Niki Parmar, et al. Atten- tion is all you need. InNeurIPS, 2017. 2, 4

    work page 2017

  31. [31]

    Perception Encoder: The best visual embeddings are not at the output of the network

    Daniel Bolya, Po-Yao Huang, Peize Sun, Jang Hyun Cho, Andrea Madotto, Chen Wei, Tengyu Ma, Jiale Zhi, Jathushan Rajasegaran, Hanoona Rasheed, et al. Perception encoder: The best visual embeddings are not at the output of the net- work.arXiv preprint arXiv:2504.13181, 2025. 2, 3

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  32. [32]

    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.arXiv preprint arXiv:2106.09685, 2021. 2, 3, 4, 7

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  33. [33]

    Cnn-generated images are surprisingly easy to spot

    Sheng-Yu Wang, Oliver Wang, Richard Zhang, Andrew Owens, and Alexei A Efros. Cnn-generated images are surprisingly easy to spot... for now. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 8695–8704, 2020. 2, 3, 5, 6

    work page 2020

  34. [34]

    Frequency-aware deepfake de- tection: Improving generalizability through frequency space domain learning

    Chuangchuang Tan, Yao Zhao, Shikui Wei, Guanghua Gu, Ping Liu, and Yunchao Wei. Frequency-aware deepfake de- tection: Improving generalizability through frequency space domain learning. InProceedings of the AAAI Conference, pages 5052–5060, 2024. 2, 5, 6

    work page 2024

  35. [35]

    Learning on gradients: Generalized arti- facts representation for gan-generated images detection

    Chuangchuang Tan, Yao Zhao, Shikui Wei, Guanghua Gu, and Yunchao Wei. Learning on gradients: Generalized arti- facts representation for gan-generated images detection. In Proceedings of the IEEE/CVF Conference on Computer Vi- sion and Pattern Recognition, pages 12105–12114, 2023. 2, 5, 6

    work page 2023

  36. [36]

    Detecting GAN generated Fake Images using Co-occurrence Matrices

    Lakshmanan Nataraj, Tajuddin Manhar Mohammed, Shiv- kumar Chandrasekaran, Arjuna Flenner, Jawadul H Bappy, Amit K Roy-Chowdhury, and BS Manjunath. Detecting gan generated fake images using co-occurrence matrices.arXiv preprint arXiv:1903.06836, 2019. 2

    work page internal anchor Pith review Pith/arXiv arXiv 1903

  37. [37]

    On the detection of synthetic images generated by diffusion mod- els

    Riccardo Corvi, Davide Cozzolino, Giada Zingarini, Gio- vanni Poggi, Koki Nagano, and Luisa Verdoliva. On the detection of synthetic images generated by diffusion mod- els. InICASSP 2023-2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 1–5. IEEE, 2023. 2, 3

    work page 2023

  38. [38]

    What makes fake images detectable? understanding prop- erties that generalize

    Lucy Chai, David Bau, Ser-Nam Lim, and Phillip Isola. What makes fake images detectable? understanding prop- erties that generalize. InEuropean conference on computer vision, pages 103–120. Springer, 2020. 2, 5, 6

    work page 2020

  39. [39]

    Thinking in frequency: Face forgery detection by min- ing frequency-aware clues

    Yuyang Qian, Guojun Yin, Lu Sheng, Zixuan Chen, and Jing Shao. Thinking in frequency: Face forgery detection by min- ing frequency-aware clues. InEuropean conference on com- puter vision, pages 86–103. Springer, 2020. 2, 5, 6

    work page 2020

  40. [40]

    Forgery-aware adaptive transformer for generalizable synthetic image detection

    Huan Liu, Zichang Tan, Chuangchuang Tan, Yunchao Wei, Jingdong Wang, and Yao Zhao. Forgery-aware adaptive transformer for generalizable synthetic image detection. In Proceedings of the IEEE/CVF Conference on Computer Vi- sion and Pattern Recognition, pages 10770–10780, 2024. 2, 5, 6

    work page 2024

  41. [41]

    Leveraging rep- resentations from intermediate encoder-blocks for synthetic image detection

    Christos Koutlis and Symeon Papadopoulos. Leveraging rep- resentations from intermediate encoder-blocks for synthetic image detection. InEuropean Conference on Computer Vi- sion, pages 394–411. Springer, 2024. 2, 5, 6

    work page 2024

  42. [42]

    Forgelens: Data- efficient forgery focus for generalizable forgery image detec- tion, 2025

    Yingjian Chen, Lei Zhang, and Yakun Niu. Forgelens: Data- efficient forgery focus for generalizable forgery image detec- tion, 2025. 3, 5, 6

    work page 2025

  43. [43]

    Effort: Efficient orthogonal mod- eling for generalizable ai-generated image detection.Pro- ceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2, 2024

    Zhiyuan Yan, Jiangming Wang, Zhendong Wang, Peng Jin, Ke-Yue Zhang, Shen Chen, Taiping Yao, Shouhong Ding, Baoyuan Wu, and Li Yuan. Effort: Efficient orthogonal mod- eling for generalizable ai-generated image detection.Pro- ceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2, 2024. 3

    work page 2024

  44. [44]

    Dire for diffusion-generated image detection

    Zhendong Wang, Jianmin Bao, Wengang Zhou, Weilun Wang, Hezhen Hu, Hong Chen, and Houqiang Li. Dire for diffusion-generated image detection. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 22445–22455, 2023. 3

    work page 2023

  45. [45]

    Aligned datasets improve detection of latent diffusion-generated images.Proceedings of the Inter- nation Conference on Learning Representations, 2025

    Anirudh Sundara Rajan, Utkarsh Ojha, Jedidiah Schloesser, and Yong Jae Lee. Aligned datasets improve detection of latent diffusion-generated images.Proceedings of the Inter- nation Conference on Learning Representations, 2025. 3

    work page 2025

  46. [46]

    Stay-positive: A case for ignoring real image features in fake image detec- tion.Proceedings of the Internation Conference on Machine Learning, 2025

    Anirudh Sundara Rajan and Yong Jae Lee. Stay-positive: A case for ignoring real image features in fake image detec- tion.Proceedings of the Internation Conference on Machine Learning, 2025. 3

    work page 2025

  47. [47]

    To- wards real-world blind face restoration with generative facial prior

    Xintao Wang, Yu Li, Honglun Zhang, and Ying Shan. To- wards real-world blind face restoration with generative facial prior. InProceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition (CVPR), pages 9168– 9178, 2021. 3, 5, 6, 7, 8

    work page 2021

  48. [48]

    Real-ESRGAN: Training real-world blind super-resolution with pure synthetic data

    Xintao Wang, Liangbin Xie, Chao Dong, and Ying Shan. Real-ESRGAN: Training real-world blind super-resolution with pure synthetic data. InProceedings of the IEEE/CVF In- ternational Conference on Computer Vision Workshops (IC- CVW), pages 1905–1914, 2021. 3, 5, 6, 7, 8

    work page 1905

  49. [49]

    FFDNet: Toward a fast and flexible solution for CNN-based im- age denoising.IEEE Transactions on Image Processing, 27(9):4608–4622, 2018

    Kai Zhang, Wangmeng Zuo, and Lei Zhang. FFDNet: Toward a fast and flexible solution for CNN-based im- age denoising.IEEE Transactions on Image Processing, 27(9):4608–4622, 2018. 3, 5, 6, 7, 8

    work page 2018

  50. [50]

    EnlightenGAN: Deep light enhancement without paired supervision.IEEE Transactions on Image Process- ing, 30:2340–2349, 2021

    Yifan Jiang, Xinyu Gong, Ding Liu, Yu Cheng, Chen Fang, Xiaohui Shen, Jianchao Yang, Pan Zhou, and Zhangyang Wang. EnlightenGAN: Deep light enhancement without paired supervision.IEEE Transactions on Image Process- ing, 30:2340–2349, 2021. 5, 6, 7, 8

    work page 2021

  51. [51]

    Towards understanding convergence and generalization of adamw.IEEE transactions on pattern analysis and machine intelligence, 46(9):6486–6493, 2024

    Pan Zhou, Xingyu Xie, Zhouchen Lin, and Shuicheng Yan. Towards understanding convergence and generalization of adamw.IEEE transactions on pattern analysis and machine intelligence, 46(9):6486–6493, 2024. 4

    work page 2024

  52. [52]

    Unpaired image-to-image translation using cycle- consistent adversarial networks

    Jun-Yan Zhu, Taesung Park, Phillip Isola, and Alexei A Efros. Unpaired image-to-image translation using cycle- consistent adversarial networks. InProceedings of the IEEE International Conference on Computer Vision, pages 2223– 2232, 2017. 5

    work page 2017

  53. [53]

    Large Scale GAN Training for High Fidelity Natural Image Synthesis

    Andrew Brock, Jeff Donahue, and Karen Simonyan. Large scale gan training for high fidelity natural image synthesis. arXiv preprint arXiv:1809.11096, 2018. 5

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  54. [54]

    A style-based generator architecture for generative adversarial networks

    Tero Karras, Samuli Laine, and Timo Aila. A style-based generator architecture for generative adversarial networks. InProceedings of the IEEE/CVF conference on computer vi- sion and pattern recognition, pages 4401–4410, 2019. 5

    work page 2019

  55. [55]

    Semantic image synthesis with spatially-adaptive nor- malization

    Taesung Park, Ming-Yu Liu, Ting-Chun Wang, and Jun-Yan Zhu. Semantic image synthesis with spatially-adaptive nor- malization. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 2337–2346,

    work page

  56. [56]

    Stargan: Unified genera- tive adversarial networks for multi-domain image-to-image translation

    Yunjey Choi, Minje Choi, Munyoung Kim, Jung-Woo Ha, Sunghun Kim, and Jaegul Choo. Stargan: Unified genera- tive adversarial networks for multi-domain image-to-image translation. InProceedings of the IEEE conference on computer vision and pattern recognition, pages 8789–8797,

    work page

  57. [57]

    Faceforen- sics++: Learning to detect manipulated facial images

    Andreas Rossler, Davide Cozzolino, Luisa Verdoliva, Chris- tian Riess, Justus Thies, and Matthias Nießner. Faceforen- sics++: Learning to detect manipulated facial images. In Proceedings of the IEEE/CVF international conference on computer vision, pages 1–11, 2019. 5

    work page 2019

  58. [58]

    Learning to see in the dark

    Chen Chen, Qifeng Chen, Jia Xu, and Vladlen Koltun. Learning to see in the dark. InProceedings of the IEEE con- ference on computer vision and pattern recognition, pages 3291–3300, 2018. 5

    work page 2018

  59. [59]

    Second-order attention network for single im- age super-resolution

    Tao Dai, Jianrui Cai, Yongbing Zhang, Shu-Tao Xia, and Lei Zhang. Second-order attention network for single im- age super-resolution. InProceedings of the IEEE/CVF con- ference on computer vision and pattern recognition, pages 11065–11074, 2019. 5

    work page 2019

  60. [60]

    Photographic image syn- thesis with cascaded refinement networks

    Qifeng Chen and Vladlen Koltun. Photographic image syn- thesis with cascaded refinement networks. InProceedings of the IEEE international conference on computer vision, pages 1511–1520, 2017. 5, 6

    work page 2017

  61. [61]

    Diverse image synthesis from semantic layouts via conditional imle

    Ke Li, Tianhao Zhang, and Jitendra Malik. Diverse image synthesis from semantic layouts via conditional imle. 2019 ieee. InCVF International Conference on Computer Vision (ICCV), pages 4219–4228, 2019. 5, 6

    work page 2019

  62. [62]

    Diffusion models beat gans on image synthesis.Advances in neural informa- tion processing systems, 34:8780–8794, 2021

    Prafulla Dhariwal and Alexander Nichol. Diffusion models beat gans on image synthesis.Advances in neural informa- tion processing systems, 34:8780–8794, 2021. 5

    work page 2021

  63. [63]

    High-resolution image synthesis with latent diffusion models

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Bj ¨orn 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. 5, 6

    work page 2022

  64. [64]

    GLIDE: Towards Photorealistic Image Generation and Editing with Text-Guided Diffusion Models

    Alex Nichol, Prafulla Dhariwal, Aditya Ramesh, Pranav Shyam, Pamela Mishkin, Bob McGrew, and Sutskever. Glide: Towards photorealistic image generation and edit- ing with text-guided diffusion models.arXiv preprint arXiv:2112.10741, 2021. 5, 6

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  65. [65]

    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 confer- ence on machine learning, pages 8821–8831. Pmlr, 2021. 5

    work page 2021

  66. [66]

    Chan, Chongyi Li, and Chen Change Loy

    Shangchen Zhou, Kelvin C.K. Chan, Chongyi Li, and Chen Change Loy. Towards robust blind face restoration with codebook lookup transformer. InAdvances in Neu- ral Information Processing Systems (NeurIPS), volume 35, pages 30599–30611, 2022. 5, 6, 7, 8

    work page 2022

  67. [67]

    Double compression detection in hevc-coded video with the same coding parameters using picture partitioning information

    Kutub Uddin, Yoonmo Yang, and Byung Tae Oh. Double compression detection in hevc-coded video with the same coding parameters using picture partitioning information. Signal Processing: Image Communication, 103:116638,

    work page

  68. [68]

    Diff- BIR: Towards blind image restoration with generative diffu- sion prior

    Xinqi Lin, Jingwen He, Ziyan Chen, Zhaoyang Lyu, Ben Fei, Bo Dai, Wanli Ouyang, Yu Qiao, and Chao Dong. Diff- BIR: Towards blind image restoration with generative diffu- sion prior. InProceedings of the European Conference on Computer Vision (ECCV), 2024. 8

    work page 2024

  69. [69]

    SDEdit: Guided image synthesis and editing with stochastic differential equa- tions

    Chenlin Meng, Yutong He, Yang Song, Jiaming Song, Jia- jun Wu, Jun-Yan Zhu, and Stefano Ermon. SDEdit: Guided image synthesis and editing with stochastic differential equa- tions. InInternational Conference on Learning Representa- tions (ICLR), 2022. 8

    work page 2022

  70. [70]

    Towards unsupervised deep image enhancement with generative adversarial network.IEEE Transactions on Image Processing, 29:9140–9151, 2020

    Zhen Ni, Wenhan Yang, Shiqi Wang, Lin Ma, and Sam Kwong. Towards unsupervised deep image enhancement with generative adversarial network.IEEE Transactions on Image Processing, 29:9140–9151, 2020. 8

    work page 2020

  71. [71]

    Anti- forensic against double jpeg compression detection using ad- versarial generative network.In Proceedings of the Korean Society of Broadcast Engineers Conference, pages 58–60,

    Kutub Uddin, Yoonmo Yang, and Byung Tae Oh. Anti- forensic against double jpeg compression detection using ad- versarial generative network.In Proceedings of the Korean Society of Broadcast Engineers Conference, pages 58–60,

    work page

  72. [72]

    Analysis of generative adversarial network targeting anti-forensic in jpeg compressed domain

    Kutub Uddin, Yoonmo Yang, and Byung Tae Oh. Analysis of generative adversarial network targeting anti-forensic in jpeg compressed domain. InInternational Workshop on Ad- vanced Imaging Technology (IWAIT) 2021, volume 11766, pages 627–631. SPIE, 2021. 9

    work page 2021

  73. [73]

    Guard: Generative unmasking and adversarial-resistant deepfake detection using multi-model knowledge distillation.Authorea Preprints, 2025

    Kutub Uddin, Nusrat Tasnim, Muhammad Saad Saeed, and Khalid Mahmood Malik. Guard: Generative unmasking and adversarial-resistant deepfake detection using multi-model knowledge distillation.Authorea Preprints, 2025. 9

    work page 2025