Pith. sign in

REVIEW 4 major objections 6 minor 71 references

Bi-Level Optimization for Self-Supervised AI-Generated Face Detection

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

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

Pith's one-line read A detector pretrained only on real face photographs can flag AI-generated faces across nine unseen generators.

desk verdict Solid engineering extension of the authors' EXIF-based detector; the cross-generator numbers are believable, but the bi-level contribution is thin and the missing variance is a real issue. read the letter →

arxiv 2507.22824 v1 pith:VOVUNUCX submitted 2025-07-30 cs.CV

Mian Zou , Nan Zhong , Baosheng Yu , Yibing Zhan , Kede Ma This is my paper
classification cs.CV
keywords AI-generatedfacedetectionself-supervisedlearningbi-leveloptimizationEXIFmetadatamanipulationone-classclassificationcross-generatorgeneralizationanomaly
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 tries to prove that AI-generated faces can be detected without ever training on a synthesized face, by steering self-supervised learning toward the right proxy. Its method, BLADES, pretrains a vision encoder on real face photographs with three families of tasks—classifying categorical EXIF camera tags, ranking ordinal EXIF tags, and detecting locally flipped or affine-warped face regions—and then uses bi-level optimization to adjust the task weights so the feature space is best aligned with coarse manipulation detection, which stands in for AI-generation detection. After pretraining, the encoder is frozen; fakes are caught either as anomalies under a Gaussian mixture model fit to real-photo features or by a small two-layer classifier trained on the frozen features. The paper reports an average cross-generator accuracy of 91.86% for the binary variant and 88.01% for the one-class variant across nine generators, beating supervised detectors that were trained on synthetic faces. If that holds, fake-face detection can be built from real photographs alone and can keep working when new generators appear.

What carries the argument

The load-bearing mechanism is the bi-level optimization of Eq. (1): $\min_\lambda \sum_{x \in B_{\mathrm{val}}} \ell_1(x; \theta^*)$ subject to $\theta^* = \arg\min_\theta \sum_{x \in B_{\mathrm{tr}}}\sum_{i=1}^{K} \lambda_i \ell_i(x;\theta)$. Here $\ell_1$ is the coarse manipulation-detection surrogate, the $\ell_i$ are the EXIF and manipulation pretext losses, and $\lambda$ is the task-weight vector. The encoder and the weights are updated alternately, with a finite-difference approximation of the mixed second-order derivative used to compute the outer-loop gradient. This converts a bag of generic self-supervised tasks into a detector-aware pretraining objective, because the weights are chosen to make the learned feature space better at separating manipulated faces from photographic faces.

What would settle it

Take the one-class variant and test it on faces produced by a new generator whose outputs reproduce photographic EXIF statistics and contain no flip-or-warp artifacts; if the detector's accuracy falls to chance while supervised detectors stay high, the surrogate-transfer assumption would be falsified.

Watch

Extended reading notes

Core claim

The central claim is that coarse-grained face manipulation detection—flagging local flips and global affine warps—is a faithful enough surrogate for AI-generated face detection that optimizing task weights against it aligns self-supervised representations with the true detection goal. BLADES never sees a synthesized face during pretraining: the inner loop minimizes a weighted sum of pretext losses on real face photographs, while the outer loop re-weights those losses so that the updated encoder reduces the manipulation-detection loss on a separate validation minibatch. The learned weights concentrate on manipulation detection and on exposure-related EXIF tags, which the paper reads as evidence that these tasks expose subtle statistical differences between photographic and generated faces. Afterwards the text encoder is discarded and the frozen vision encoder supports both GMM anomaly detection and a two-layer perceptron classifier. The paper reports this recipe surpasses ten competing detectors on average over StyleGAN2, VQGAN, LDM, DDIM, SDv2.1, FreeDoM, HPS, Midjourney, and SDXL, and that the representations transfer from CelebA-HQ photographs to FFHQ photographs and FFHQ-trained generators.

Load-bearing premise

The load-bearing premise is that detecting locally flipped or affine-warped face regions is a reliable stand-in for detecting AI-generated faces; if that surrogate stops transferring to new generators, the whole bi-level alignment and the one-class results lose their target.

Editorial extensions

If this is right

  • A frozen encoder that has never seen AI-generated faces can act as a reusable feature extractor, so a new generator can be screened immediately without retraining.
  • The binary detector is light enough for deployment: the paper reports about 26 million parameters, roughly 4.1 billion FLOPs, and 0.005 seconds per image, which is 8x fewer parameters, 12x fewer FLOPs, and 20x faster inference than the leading supervised baseline.
  • Cross-dataset results on FFHQ indicate that the representations transfer across photographic domains, not only across generators.
  • Under mild JPEG compression the method retains higher accuracy than FatFormer on most generators, though the paper notes a visible drop in overall accuracy and names robustness to post-processing as an open problem.
  • The paper's own discussion anticipates that broader natural-scene forgery detection will require new surrogate tasks beyond face manipulations, such as semantic consistency or temporal coherence.

Reading between the lines

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

  • A testable extension is to replace the coarse manipulation surrogate with a domain-appropriate proxy—scene composition for natural images or temporal coherence for video—and check whether bi-level weighting rediscovers the most useful self-supervised tasks automatically.
  • The high learned weights on exposure-related EXIF tags suggest the detector may partly rely on camera-pipeline statistics that generators do not reproduce; stripping or randomizing EXIF metadata during pretraining would test this directly.
  • Because the pipeline needs no synthetic training data, it could serve as a continuously updated forensic baseline in settings where new generators appear faster than labeled fake data can be collected.
  • The outer-loop scheme is effectively an automatic curriculum for self-supervised tasks, so the same alignment idea could apply to other detection problems where a cheap surrogate is easier to label than the actual target.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper introduces BLADES, a self-supervised pretraining method for AI-generated face detection. A joint vision-text encoder is pretrained only on face photographs using four pretext tasks: coarse-grained manipulation detection (the surrogate primary task), categorical EXIF tag classification, ordinal EXIF tag ranking, and fine-grained manipulation detection. A bi-level optimization loop updates the encoder parameters in the inner loop and the task weights in the outer loop, where the outer objective is the validation loss of the coarse manipulation task. After pretraining, the encoder is frozen and used either with a GMM anomaly detector (one-class) or a small two-layer perceptron classifier (binary). Experiments across nine generators and two photographic datasets report substantial gains over prior detectors, along with ablations, JPEG robustness, and efficiency comparisons.

Significance. If the empirical claims hold, BLADES is a meaningful contribution: a single frozen encoder trained without synthetic faces in its one-class variant transfers across GAN and diffusion generators and is substantially more efficient than a strong Transformer-based competitor. The paper has notable strengths: a broad comparison across nine generators, cross-dataset evaluation, ablations of the model structure, training strategy, primary task, and pretext composition, a GitHub code release, and an explicit robustness analysis. The main caveats are that the cross-generator mechanism rests on an unvalidated surrogate assumption, the reported margins are single point estimates without uncertainty, and the binary-classification protocol is underspecified. Conditional on the additional evidence requested below, the results would be of interest to the forensics and self-supervised learning communities.

major comments (4)
  1. [Sec. 3.2–3.3, Eq. (1a)] The outer-loop objective in Eq. (1a) is the same as the inner-loop pretext loss ℓ1: coarse-grained face manipulation detection on locally flipped or affine-warped photographs. Because ℓ1 is also one of the terms in Eq. (1b), the optimizer can succeed simply by up-weighting this proxy; it never observes an AI-generated face in either loop. The paper's cross-generator claim therefore depends entirely on an empirical equivalence between these manipulations and faces from unseen generators. Table 7c compares alternative primary tasks but is evaluated on the same nine generators and only on the two manipulation families used in pretraining, and Section 5 explicitly concedes that “new surrogate tasks beyond face manipulations” are needed. I request a direct test of proxy transferability, e.g., evaluating BLADES on manipulation families not used in pretraining or on held-out generators after the outer loop has been optimized only on the proxy, and reporting the result alongside Table 7c.
  2. [Sec. 3.4 and Table 1] The evaluation protocol for BLADES-BC is underspecified. Section 4.2 states that detectors are trained with 25,000 CelebA-HQ photographs and synthetic images from a single generator, but it is not stated which generator supplies the positive examples for the two-layer perceptron in Sec. 3.4, nor how many examples are used. If the perceptron is trained on all nine generators, the “unseen generators” claim for BLADES-BC in Table 1 is not valid; if it is trained on one generator, that generator and the number of samples should be stated and matched across baselines. In addition, the BC variant uses low-likelihood photographic pseudo-outliers during classifier training, so it is not purely self-supervised; the distinction in the abstract between the one-class and binary settings should be clarified.
  3. [Tables 1–7] All reported numbers are single point estimates. For example, Table 1 reports 91.86% average accuracy for BLADES-BC versus 89.68% for FatFormer, and Table 7b reports that bi-level optimization improves mAcc from 86.20% to 88.01%; without repeated runs, confidence intervals, or paired significance tests, the abstract's claim that BLADES “significantly outperform[s] existing approaches” is not supported. I request mean and standard deviation over at least three seeds for the main tables and paired tests where the same test sets are used.
  4. [Supplementary Eqs. (14)–(18)] The bi-level method is implemented with a one-step inner update θ′ = θ − α∇θ ℓpre (Eq. 11) and a central finite-difference approximation of the hypergradient, rather than the exact inner-loop minimizer θ⋆ in Eq. (1b). The manuscript presents no evidence that this approximation preserves the intended bi-level solution, and no sensitivity analysis over the inner-loop learning rate, number of inner steps, or finite-difference step ϵ. Since the task-weighting mechanism is the paper's main novelty, I ask for a sensitivity study on a subset of the data, or at least a comparison between the one-step and multi-step approximations.
minor comments (6)
  1. [Fig. 4] The learned task weights in Fig. 4 are rendered as Unicode escape sequences (e.g., “/uni...”), so neither the task names nor the weight values are readable; please replace with a legible figure and caption.
  2. [Sec. 4.3 and Table 5] The sentence “accuracy drops from 91.86% to 78.87%” does not match Table 5, which reports AUC; clarify which metric is being reported.
  3. [Table 2] Table 2 reports an F-score without defining the formula or giving sample sizes per generator; please add these details.
  4. [Sec. 3.2 and Eq. (2)] The text encoder and vision encoder are denoted by visually similar symbols (fφ and fϕ); use distinct names such as f_V and f_T to avoid ambiguity.
  5. [Sec. 4.1] The description “filtering nine informative tags with 130,000 faces remaining” is ambiguous; clarify whether 130,000 is the number of images that contain all nine selected tags.
  6. [Table 1 note] The note under Table 1 says “our OC variant” is not trained on synthetic faces but does not state whether BLADES-BC's classifier saw synthetic faces; clarify this in the caption.

Circularity Check

1 steps flagged · score 2.0 of 10

Main detection results are independently benchmarked against held-out generators; the only self-referential element is the Fig. 4 task-relevance claim, since the outer loop minimizes the same ℓ1 that is also an inner-loop task.

  1. self definitional [Section 3.1 Eq. (1a)-(1b); Section 3.3 Eq. (12); Section 4.3 Task Relationship Analysis / Fig. 4]
    "The outer loop then optimizes the relative weights of these pretext tasks to enhance the coarse-grained detection of manipulated faces, serving as a proxy task for identifying AI-generated faces. ... λ ← λ − β∇λ Σ_{x∈Bval} ℓ1(x; θ′(λ)) ... First, coarse-grained face manipulation detection—used as the surrogate primary task—receives the highest weight. This confirms its pivotal role in aligning the self-supervised feature space with the downstream goal of AI-generated face detection."

    Task 1 (coarse-grained face manipulation detection) defines ℓ1, which is one of the inner-loop pretext losses in Eq. (1b) and also the outer-loop validation objective in Eqs. (1a) and (12). The learned weights λ are therefore optimized to minimize ℓ1 itself. The observation that Task 1 receives the highest weight in Fig. 4 is a consequence of this optimization setup, not an independent finding that coarse manipulation is the most relevant surrogate for AI-generated faces. 'Alignment with AI-generated face detection' is equated by definition with minimizing ℓ1 on locally flipped/affine-warped photographs; no AI-generated face appears in the outer loop.

full rationale

The core empirical claims are self-contained: BLADES-OC and BLADES-BC are evaluated on nine held-out generators and against ten baselines (Table 1), with additional cross-dataset (Table 3) and separability (Table 4) experiments, and no AI-generated face is used in the outer-loop objective. The cross-generator accuracy therefore does not reduce to the fitted task weights. The one self-referential element is that Task 1's loss ℓ1 is simultaneously an inner-loop pretext loss and the outer-loop validation objective (Eqs. (1b) and (12)). Consequently, the high weight assigned to coarse-grained manipulation detection in Fig. 4 is partly by construction, and the statement that this 'confirms its pivotal role' should be read as a property of the optimization target rather than independent evidence about AI-generated face relevance. The paper itself concedes in Section 5 that new surrogate tasks beyond face manipulations are needed. The authors' prior work [66] supplies the manipulation proxy and the GMM anomaly-detection recipe, but the present paper re-implements that baseline and outperforms it on the same benchmarks, so the self-citation is not used as unverified load-bearing support. Overall: no load-bearing circularity in the main results; one minor self-referential interpretation.

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

The method's load-bearing assumptions are domain transfer assumptions: the manipulation proxy, EXIF cross-dataset transfer, and the one-step hypergradient approximation. No new physical entities, forces, or conserved quantities are introduced, and the pseudo-outlier synthesis in the binary classifier is borrowed from VOS [11]. The fitted meta-parameters (λ, γ, GMM components, anomaly threshold, EXIF tag set) are modest in number and standard for this type of method.

free parameters (5)
  • Task weight vector λ = learned; not reported numerically
    Optimized by the outer loop on validation minibatches via Eq. (1a); this is the central mechanism of the method and is fitted to data rather than derived.
  • Focal loss exponent γ = 2
    Set by hand following Lin et al. [30]; affects the categorical EXIF loss in Eq. (5).
  • GMM component count = 10
    Set by hand for the one-class detector in Section 4.1; no sensitivity analysis is provided.
  • Anomaly threshold = 5th percentile of training features
    Fixes the false alarm rate for the one-class accuracy numbers; other thresholds would change reported Acc values.
  • EXIF tag selection = 9 tags from 19 candidates
    Chosen by three empirical filtering criteria in Appendix B; other tag sets could change the learned representation.
assumptions (4)
  • domain assumption Coarse-grained face manipulation detection (local flips and global affine warps) is a valid proxy for AI-generated face detection.
    Task 1 in Section 3.2 defines this surrogate; Eq. (1a) makes its validation loss the outer objective. No AI-generated face is seen in pretraining, so transfer to unseen generators depends on this proxy.
  • domain assumption EXIF metadata statistics from the FDF collection transfer to CelebA-HQ and FFHQ photographic domains.
    Section 4.1 pretrains on FDF photos but tests on CelebA-HQ and FFHQ; the cross-dataset results in Table 3 provide empirical, not first-principles, support.
  • ad hoc to paper The one-step inner update and finite-difference hypergradient approximation (supplementary Eqs. 14-18) faithfully represent the true bi-level solution.
    This is the standard DARTS/Auto-Lambda style approximation; if it diverges from the exact bilevel optimum, the learned weights may not reflect true task benefit.
  • standard math Thurstone's model with a normal CDF (Eq. 8) links discretized EXIF levels to pairwise ordering probabilities.
    Standard comparative-judgment model used to define the ordinal EXIF ranking loss; an assumption about the noise distribution.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Bi-Level Optimization for Self-Supervised AI-Generated Face Detection." pith.science (2026). https://pith.science/paper/VOVUNUCX

@misc{pith2026250722824,
  author       = {Pith},
  title        = {Pith review of: Bi-Level Optimization for Self-Supervised AI-Generated Face Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VOVUNUCX}},
  note         = {Machine review of arXiv:2507.22824}
}
read the original abstract

AI-generated face detectors trained via supervised learning typically rely on synthesized images from specific generators, limiting their generalization to emerging generative techniques. To overcome this limitation, we introduce a self-supervised method based on bi-level optimization. In the inner loop, we pretrain a vision encoder only on photographic face images using a set of linearly weighted pretext tasks: classification of categorical exchangeable image file format (EXIF) tags, ranking of ordinal EXIF tags, and detection of artificial face manipulations. The outer loop then optimizes the relative weights of these pretext tasks to enhance the coarse-grained detection of manipulated faces, serving as a proxy task for identifying AI-generated faces. In doing so, it aligns self-supervised learning more closely with the ultimate goal of AI-generated face detection. Once pretrained, the encoder remains fixed, and AI-generated faces are detected either as anomalies under a Gaussian mixture model fitted to photographic face features or by a lightweight two-layer perceptron serving as a binary classifier. Extensive experiments demonstrate that our detectors significantly outperform existing approaches in both one-class and binary classification settings, exhibiting strong generalization to unseen generators.

Figures

Figures reproduced from arXiv: 2507.22824 by the authors.

Figure 1
Figure 1. System diagram of the proposed BLADES for AI-generated face detection. [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Visualization of artificially manipulated face pho [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. t-SNE visualizations [50] of learned features of photographic (yellow) and AI-generated (blue) face images across five generators. our BLADES-OC, trained exclusively on face photographs, surpasses most supervised detectors despite never being ex￾posed to AI-generated faces. This highlights the effective￾ness of our bi-level optimization strategy in aligning pre￾text tasks with the downstream detection objective. In … view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Learned loss weights across pretext tasks, where higher [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Representative face images used in cross-generator evaluation. (Part 1 of 2). [PITH_FULL_IMAGE:figures/full_fig_p013_5.png]
Figure 5
Figure 5. Figure 5: Representative face images used in cross-generator evaluation. (Part 2 of 2). [PITH_FULL_IMAGE:figures/full_fig_p014_5.png]
Figure 6
Figure 6. Figure 6: Representative face images used in cross-dataset evaluation. [PITH_FULL_IMAGE:figures/full_fig_p015_6.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

71 extracted references · 65 canonical work pages

  1. [1]

    https://www.midjourney.com

    Midjourney. https://www.midjourney.com. Accessed: Jul 30, 2025

    work page 2025

  2. [2]

    A geometric and photometric exploration of GAN and diffusion synthesized faces

    Maty \'a s Boh \'a c ek and Hany Farid. A geometric and photometric exploration of GAN and diffusion synthesized faces. In CVPRW, pages 874--883, 2023

    work page 2023

  3. [3]

    End-to-end reconstruction-classification learning for face forgery detection

    Junyi Cao, Chao Ma, Taiping Yao, Shen Chen, Shouhong Ding, and Xiaokang Yang. End-to-end reconstruction-classification learning for face forgery detection. In CVPR, pages 4113--4122, 2022

    work page 2022

  4. [4]

    Structure-aware protein self-supervised learning

    Can Chen, Jingbo Zhou, Fan Wang, Xue Liu, and Dejing Dou. Structure-aware protein self-supervised learning. arXiv preprint arXiv:2022.04213v4, 2022

    work page arXiv 2022

  5. [5]

    DiffusionFace : Towards a comprehensive dataset for diffusion-based face forgery analysis

    Zhongxi Chen, Ke Sun, Ziyin Zhou, Xianming Lin, Xiaoshuai Sun, Liujuan Cao, and Rongrong Ji. DiffusionFace : Towards a comprehensive dataset for diffusion-based face forgery analysis. arXiv preprint arXiv:2403.18471, 2024

    arXiv 2024

  6. [6]

    Diffusion facial forgery detection

    Harry Cheng, Yangyang Guo, Tianyi Wang, Liqiang Nie, and Mohan Kankanhalli. Diffusion facial forgery detection. In ACM MM, pages 5939--5948, 2024

    work page 2024

  7. [7]

    Intriguing properties of synthetic images: From generative adversarial networks to diffusion models

    Riccardo Corvi, Davide Cozzolino, Giovanni Poggi, Koki Nagano, and Luisa Verdoliva. Intriguing properties of synthetic images: From generative adversarial networks to diffusion models. In CVPRW, pages 973--982, 2023

    work page 2023

  8. [8]

    Zero-shot detection of AI -generated images

    Davide Cozzolino, Giovanni Poggi, Matthias Nießner, and Luisa Verdoliva. Zero-shot detection of AI -generated images. In ECCV, pages 54--72, 2024

    work page 2024

Show all 71 references
  1. [9]

    Foundations of Bilevel Programming

    Stephan Dempe. Foundations of Bilevel Programming. Springer Science & Business Media, 2002

    2002

  2. [10]

    Think twice before detecting GAN -generated fake images from their spectral domain imprints

    Chengdong Dong, Ajay Kumar, and Eryun Liu. Think twice before detecting GAN -generated fake images from their spectral domain imprints. In CVPR, pages 7865--7874, 2022

    2022

  3. [11]

    VOS : Learning what you don't know by virtual outlier synthesis

    Xuefeng Du, Zhaoning Wang, Mu Cai, and Yixuan Li. VOS : Learning what you don't know by virtual outlier synthesis. In ICLR, pages 1--21, 2022

    2022

  4. [12]

    Watch your up-convolution: CNN based generative deep neural networks are failing to reproduce spectral distributions

    Ricard Durall, Margret Keuper, and Janis Keuper. Watch your up-convolution: CNN based generative deep neural networks are failing to reproduce spectral distributions. In CVPR, pages 7890--7899, 2020

    2020

  5. [13]

    Taming T ransformers for high-resolution image synthesis

    Patrick Esser, Robin Rombach, and Bjorn Ommer. Taming T ransformers for high-resolution image synthesis. In CVPR, pages 1 2873 --1 2883 , 2021

    2021

  6. [14]

    Perceptual quality assessment of smartphone photography

    Yuming Fang, Hanwei Zhu, Yan Zeng, Kede Ma, and Zhou Wang. Perceptual quality assessment of smartphone photography. In CVPR, pages 3677--3686, 2020

    2020

  7. [15]

    Leveraging frequency analysis for deep fake image recognition

    Joel Frank, Thorsten Eisenhofer, Lea Sch \"o nherr, Asja Fischer, Dorothea Kolossa, and Thorsten Holz. Leveraging frequency analysis for deep fake image recognition. In ICML, pages 3247--3258, 2020

    2020

  8. [16]

    Unsupervised representation learning by predicting image rotations

    Spyros Gidaris, Praveer Singh, and Nikos Komodakis. Unsupervised representation learning by predicting image rotations. arXiv preprint arXiv:1803.07728, 2018

    2018 arXiv

  9. [17]

    Scaling and benchmarking self-supervised visual representation learning

    Priya Goyal, Dhruv Mahajan, Abhinav Gupta, and Ishan Misra. Scaling and benchmarking self-supervised visual representation learning. In ICCV, pages 6391--6400, 2019

    2019

  10. [18]

    Eyes tell all: Irregular pupil shapes reveal GAN -generated faces

    Hui Guo, Shu Hu, Xin Wang, Ming-Ching Chang, and Siwei Lyu. Eyes tell all: Irregular pupil shapes reveal GAN -generated faces. In ICASSP, pages 2904--2908, 2022

    2022

  11. [19]

    Douglas M. Hawkins. Identification of Outliers. Springer, 1980

    1980

  12. [20]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In CVPR, pages 770--778, 2016

    2016

  13. [21]

    Masked autoencoders are scalable vision learners

    Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Doll \'a r, and Ross Girshick. Masked autoencoders are scalable vision learners. In CVPR, pages 1 6000 --1 6009 , 2022

    2022

  14. [22]

    Exposing GAN -generated faces using inconsistent corneal specular highlights

    Shu Hu, Yuezun Li, and Siwei Lyu. Exposing GAN -generated faces using inconsistent corneal specular highlights. In ICASSP, pages 2500--2504, 2021

    2021

  15. [23]

    Minyoung Huh, Andrew Liu, Andrew Owens, and Alexei A. Efros. Fighting fake news: Image splice detection via learned self-consistency. In ECCV, pages 101--117, 2018

    2018

  16. [24]

    DeepPrivacy : A generative adversarial network for face anonymization

    H kon Hukkel s, Rudolf Mester, and Frank Lindseth. DeepPrivacy : A generative adversarial network for face anonymization. In ISVC, pages 565--578, 2019

    2019

  17. [25]

    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. In ICLR, pages 1--26, 2018

    2018

  18. [26]

    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. In CVPR, pages 4401--4410, 2019

    2019

  19. [27]

    Analyzing and improving the image quality of StyleGAN

    Tero Karras, Samuli Laine, Miika Aittala, Janne Hellsten, Jaakko Lehtinen, and Timo Aila. Analyzing and improving the image quality of StyleGAN . In CVPR, pages 8110--8119, 2020

    2020

  20. [28]

    Adam: A method for stochastic optimization

    Diederik Kingma and Jimmy Ba. Adam: A method for stochastic optimization. In ICLR, pages 1--15, 2015

    2015

  21. [29]

    Align D et: Aligning pre-training and fine-tuning in object detection

    Ming Li, Jie Wu, Xionghui Wang, Chen Chen, Jie Qin, Xuefeng Xiao, Rui Wang, Min Zheng, and Xin Pan. Align D et: Aligning pre-training and fine-tuning in object detection. In ICCV, pages 6866--6876, 2023

    2023

  22. [30]

    Focal loss for dense object detection

    Tsung-Yi Lin, Priya Goyal, Ross Girshick, Kaiming He, and Piotr Doll \'a r. Focal loss for dense object detection. In ICCV, pages 2980--2988, 2017

    2017

  23. [31]

    Detecting generated images by real images

    Bo Liu, Fan Yang, Xiuli Bi, Bin Xiao, Weisheng Li, and Xinbo Gao. Detecting generated images by real images. In ECCV, pages 95--110, 2022 a

    2022

  24. [32]

    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 CVPR, pages 1 0770 --1 0780 , 2024

    2024

  25. [33]

    Davison, and Edward Johns

    Shikun Liu, Stephen James, Andrew J. Davison, and Edward Johns. Auto- L ambda: D isentangling dynamic task relationships. arXiv preprint arXiv:2202.03091, 2022 b

    2022 arXiv

  26. [34]

    Zhengzhe Liu, Xiaojuan Qi, and Philip H.S. Torr. Global texture enhancement for fake face detection in the wild. In CVPR, pages 8060--8069, 2020

    2020

  27. [35]

    Decoupled weight decay regularization

    Ilya Loshchilov and Frank Hutter. Decoupled weight decay regularization. In ICLR, pages 1--10, 2019

    2019

  28. [36]

    LaRE ^2 : Latent reconstruction error based method for diffusion-generated image detection

    Yunpeng Luo, Junlong Du, Ke Yan, and Shouhong Ding. LaRE ^2 : Latent reconstruction error based method for diffusion-generated image detection. In CVPR, pages 1 7006 --1 7015 , 2024

    2024

  29. [37]

    Nightingale and Hany Farid

    Sophie J. Nightingale and Hany Farid. AI -synthesized faces are indistinguishable from real faces and more trustworthy. PNAS, 119 0 (8): 0 1--3, 2022

    2022

  30. [38]

    Towards universal fake image detectors that generalize across generative models

    Utkarsh Ojha, Yuheng Li, and Yong Jae Lee. Towards universal fake image detectors that generalize across generative models. In CVPR, pages 24480--24489, 2023

    2023

  31. [39]

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

    Dustin Podell, Zion English, Kyle Lacey, Andreas Blattmann, Tim Dockhorn, Jonas M \"u ller, Joe Penna, and Robin Rombach. SDXL : Improving latent diffusion models for high-resolution image synthesis. In ICLR, pages 1--13, 2024

    2024

  32. [40]

    Language models are unsupervised multitask learners

    Alec Radford, Jeffrey Wu, Rewon Child, David Luan, Dario Amodei, and Ilya Sutskever. Language models are unsupervised multitask learners. https://openai.com/index/better-language-models/. Accessed: Jul 30, 2025

    2025

  33. [41]

    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, Gretchen Krueger, and Ilya Sutskever. Learning transferable visual models from natural language supervision. In ICML, pages 8748-...

    2021

  34. [42]

    Reed, Xiangyu Yue, Ani Nrusimha, Sayna Ebrahimi, Vivek Vijaykumar, Richard Mao, Bo Li, Shanghang Zhang, Devin Guillory, Sean Metzger, Kurt Keutzer, and Trevor Darrell

    Colorado J. Reed, Xiangyu Yue, Ani Nrusimha, Sayna Ebrahimi, Vivek Vijaykumar, Richard Mao, Bo Li, Shanghang Zhang, Devin Guillory, Sean Metzger, Kurt Keutzer, and Trevor Darrell. Self-supervised pretraining improves self-supervised pretraining. In WACV, pages 2584--2594, 2022

    2022

  35. [43]

    AEROBLADE : Training-free detection of latent diffusion images using autoencoder reconstruction error

    Jonas Ricker, Denis Lukovnikov, and Asja Fischer. AEROBLADE : Training-free detection of latent diffusion images using autoencoder reconstruction error. In CVPR, pages 9130--9140, 2024

    2024

  36. [44]

    High-resolution image synthesis with latent diffusion models

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Bj \"o rn Ommer. High-resolution image synthesis with latent diffusion models. In CVPR, pages 1 0684 --1 0695 , 2022

    2022

  37. [45]

    Bi-level finetuning with task-dependent similarity structure for low-resource training

    Sai Ashish Somayajula, Lifeng Jin, Linfeng Song, Haitao Mi, and Dong Yu. Bi-level finetuning with task-dependent similarity structure for low-resource training. In ACL Findings, pages 8569--8588, 2023

    2023

  38. [46]

    Denoising diffusion implicit models

    Jiaming Song, Chenlin Meng, and Stefano Ermon. Denoising diffusion implicit models. In ICLR, pages 1--20, 2021

    2021

  39. [47]

    Learning on gradients: Generalized artifacts representation for GAN -generated images detection

    Chuangchuang Tan, Yao Zhao, Shikui Wei, Guanghua Gu, and Yunchao Wei. Learning on gradients: Generalized artifacts representation for GAN -generated images detection. In CVPR, pages 1 2105 --1 2114 , 2023

    2023

  40. [48]

    Thurstone

    Louis L. Thurstone. A law of comparative judgment. Psychol. Rev., 34: 0 273--286, 1927

    1927

  41. [49]

    F R ank: A ranking method with fidelity loss

    Ming-Feng Tsai, Tie-Yan Liu, Tao Qin, Hsin-Hsi Chen, and Wei-Ying Ma. F R ank: A ranking method with fidelity loss. In ACM SIGIR, pages 383--390, 2007

    2007

  42. [50]

    Visualizing data using t- SNE

    Laurens van der Maaten and Geoffrey Hinton. Visualizing data using t- SNE . JMLR, 9 0 (86): 0 2579--2605, 2008

    2008

  43. [51]

    SGDR : Stochastic gradient descent with warm restarts

    Ilya 0mm Loshchilov and Frank Hutter. SGDR : Stochastic gradient descent with warm restarts. In ICLR, pages 1--13, 2017

    2017

  44. [52]

    Sheng-Yu Wang, Oliver Wang, Richard Zhang, Andrew Owens, and Alexei A. Efros. CNN -generated images are surprisingly easy to spot...for now. In CVPR, pages 8695--8704, 2020

    2020

  45. [53]

    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. In ICCV, pages 2 2445 --2 2455 , 2023

    2023

  46. [54]

    Noisy T une: A little noise can help you finetune pretrained language models better

    Chuhan Wu, Fangzhao Wu, Tao Qi, Yongfeng Huang, and Xing Xie. Noisy T une: A little noise can help you finetune pretrained language models better. arXiv preprint arXiv:2202.12024, 2022

    2022 arXiv

  47. [55]

    Better aligning text-to-image models with human preference

    Xiaoshi Wu, Keqiang Sun, Feng Zhu, Rui Zhao, and Hongsheng Li. Better aligning text-to-image models with human preference. In ICCV, pages 2096--2105, 2023

    2023

  48. [56]

    Instance localization for self-supervised detection pretraining

    Ceyuan Yang, Zhirong Wu, Bolei Zhou, and Stephen Lin. Instance localization for self-supervised detection pretraining. In CVPR, pages 3987--3996, 2021

    2021

  49. [57]

    Exposing Deep Fakes using inconsistent head poses

    Xin Yang, Yuezun Li, and Siwei Lyu. Exposing Deep Fakes using inconsistent head poses. In ICASSP, pages 8261--8265, 2019

    2019

  50. [58]

    FreeDoM : Training-free energy-guided conditional diffusion model

    Jiwen Yu, Yinhuai Wang, Chen Zhao, Bernard Ghanem, and Jian Zhang. FreeDoM : Training-free energy-guided conditional diffusion model. In ICCV, pages 2 3174 --2 3184 , 2023

    2023

  51. [59]

    BiSSL : Bilevel optimization for self-supervised pre-training and fine-tuning

    Gustav Wagner Zakarias, Lars Kai Hansen, and Zheng-Hua Tan. BiSSL : Bilevel optimization for self-supervised pre-training and fine-tuning. arXiv preprint arXiv:2410.02387, 2024

    2024

  52. [60]

    A flexible new technique for camera calibration

    Zhengyou Zhang. A flexible new technique for camera calibration. IEEE TPAMI, 22 0 (11): 0 1330--1334, 2000

    2000

  53. [61]

    EXIF as language: Learning cross-modal associations between images and camera metadata

    Chenhao Zheng, Ayush Shrivastava, and Andrew Owens. EXIF as language: Learning cross-modal associations between images and camera metadata. In CVPR, pages 6945--6956, 2023

    2023

  54. [62]

    General facial representation learning in a visual-linguistic manner

    Yinglin Zheng, Hao Yang, Ting Zhang, Jianmin Bao, Dongdong Chen, Yangyu Huang, Lu Yuan, Dong Chen, Ming Zeng, and Fang Wen. General facial representation learning in a visual-linguistic manner. In CVPR, pages 1 8697 --1 8709 , 2022

    2022

  55. [63]

    Adrian Ziegler and Yuki M. Asano. Self-supervised learning of object parts for semantic segmentation. In CVPR, pages 1 4502 --1 4511 , 2022

    2022

  56. [64]

    Semantics-oriented multitask learning for DeepFake detection: A joint embedding approach

    Mian Zou, Baosheng Yu, Yibing Zhan, Siwei Lyu, and Kede Ma. Semantics-oriented multitask learning for DeepFake detection: A joint embedding approach. IEEE TCSVT, to appear, 2025

    2025

  57. [65]

    Semantic contextualization of face forgery: A new definition, dataset, and detection method

    Mian Zou, Baosheng Yu, Yibing Zhan, Siwei Lyu, and Kede Ma. Semantic contextualization of face forgery: A new definition, dataset, and detection method. IEEE TIFS, 20: 0 4512--4524, 2025 a

    2025

  58. [66]

    Self-supervised learning for detecting AI -generated faces as anomalies

    Mian Zou, Baosheng Yu, Yibing Zhan, and Kede Ma. Self-supervised learning for detecting AI -generated faces as anomalies. In ICASSP, pages 1--5, 2025 b

    2025

  59. [67]

    Model-agnostic meta-learning for fast adaptation of deep networks

    Chelsea Finn, Pieter Abbeel, and Sergey Levine. Model-agnostic meta-learning for fast adaptation of deep networks. in ICML, pages 1126--1135, 2017

    2017

  60. [68]

    DARTS: Differentiable architecture search

    Hanxiao Liu, Karen Simonyan, and Yiming Yang. DARTS: Differentiable architecture search. arXiv preprint arXiv:1806.09055, 2018

    2018 arXiv

  61. [69]

    Williamson, Alex Smola, John Shawe-Taylor, and John Platt

    Bernhard Sch \"o lkopf, Robert C. Williamson, Alex Smola, John Shawe-Taylor, and John Platt. Support vector method for novelty detection. In NeurIPS, pages 582--588, 1999

    1999

  62. [70]

    Efros, Eli Shechtman, and Oliver Wang

    Richard Zhang, Phillip Isola, Alexei A. Efros, Eli Shechtman, and Oliver Wang. The unreasonable effectiveness of deep features as a perceptual metric. In CVPR, pages 586--595, 2018

    2018

  63. [71]

    write newline

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

Pith tools

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