Pith. sign in

REVIEW 3 major objections 5 minor 36 references

Knowledge-Guided Prompt Learning for Deepfake Facial Image Detection

T0 review · 3 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read A frozen CLIP with 768 trainable parameters detects deepfake faces at 0.911 average AUC, beating fully tuned state-of-the-art detectors on DeepFakeFaceForensics.

desk verdict Clever idea, but the SOTA claim leans on a transductive test-time tuning stage that makes the comparison protocol-asymmetric. read the letter →

arxiv 2501.00700 v2 pith:NQA34KMB submitted 2025-01-01 cs.CV

Hao Wang , Cheng Deng , Zhidong Zhao This is my paper · ORCID
classification cs.CV
keywords deepfakedetectionpromptlearningCLIPtest-timeadaptationlargelanguagemodeldomainshiftfacialimageforensics
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

This paper tries to establish that deepfake face detection improves when a frozen CLIP vision-language model is steered by two moves: prompts built from forgery-related concepts retrieved from a large language model, and a second round of prompt tuning on unlabeled test images using pseudo-labels the model assigns by its own confidence. The authors argue that prior detectors ignore both expert knowledge and the category-level domain shift between training images (natural objects such as horses and chairs) and test images (fine-grained human faces), and that these gaps can be closed without touching the backbone weights. If true, the result matters because it reports 0.911 average AUC and 0.774 overall accuracy on DeepFakeFaceForensics, an absolute AUC gain of 0.098 over the previous state of the art, using only 768 trainable parameters.

What carries the argument

The central objects are the two prompt templates $P_{real}$ and $P_{fake}^{i}$, each a learnable token sequence followed by a class description: $[real]$ for pristine images, or one of the LLM-retrieved concepts $\phi_i$ from the set $\Phi = \{\text{fake}, \text{blurred}, \dots, \text{unrealistic}\}$ for fakes. The frozen CLIP text encoder $G$ maps these to prototypes $G_{real}$ and $G_{fake} = \frac{1}{|\Phi|}\sum_{i} G(P_{fake}^{i})$, while the frozen image encoder $F$ yields cosine similarities $S_{real}$ and $S_{fake}$; a temperature-scaled softmax converts them into the probabilities $\delta_{real}$ and $\delta_{fake}$ used in both the training loss and the test-time selection rule (Eqs. 5-13). The machinery's job is to turn a frozen vision-language model into a detector whose only movable parts are the prompts, so that detection can be re-adapted at inference time to the unlabeled test distribution.

What would settle it

Run the same test-time prompt tuning with the pseudo-labels randomly permuted among the same selected samples: if average AUC stays at 0.911 and overall accuracy at 0.774, the gain is not coming from learning real-versus-fake structure. Alternatively, give the comparison baselines the same transductive access to the unlabeled test batch (for example, entropy minimization or the same confident-sample selection without the LLM concepts) and check whether the 0.098 absolute AUC gap over GLFF persists; if it does, the protocol, not the knowledge guidance, carries the headline number.

Watch

Extended reading notes

Core claim

The discovery, on the paper's own terms, is that a single learnable prompt per class, seeded with LLM-retrieved forgery concepts, plus a transductive prompt-tuning stage on pseudo-labeled test data, yields top results on deepfake facial image detection. The fake-class prompt is not the bare word 'fake': it is an averaged text prototype over concepts such as 'blurred', 'unnatural' and 'inconsistent' collected by querying GPT-4, which the authors treat as expert-level prior knowledge that also makes the prompts interpretable. The test-time stage selects up to 128 high-confidence samples per class with thresholds $T_{real}=0.999$ and $T_{fake}=0.5$, assigns pseudo-labels, and continues optimizing the same prompts; the ablation attributes a large overall-accuracy jump (from 0.598 without test-time tuning to 0.774 with it) to this step.

Load-bearing premise

The design assumes that tuning prompts on test images the model itself marks as high-confidence is a valid inference-time step, and that the gain over inductive baselines is a fair measure of detection quality; if the self-generated pseudo-labels simply agree with the model's own mistakes, the reported improvement could reflect self-confirmation rather than genuine domain-shift correction.

Editorial extensions

If this is right

  • A frozen CLIP ViT-L/14 with 768 trainable prompt parameters would outperform fully fine-tuned detectors, cutting deployment cost by orders of magnitude (0.911 vs. 0.813 average AUC against GLFF's 26.8M trained parameters).
  • Test-time prompt tuning on self-selected, pseudo-labeled test images would be a practical way to adapt a detector to a target domain at inference time without ground-truth labels.
  • LLM-retrieved forensic concepts would transfer across unseen generative models, since the presented results cover GAN-based, transformer-based and diffusion-based generators.
  • The ablations imply the two components contribute differently: knowledge-guided prompts lift both AUC and overall accuracy, while test-time tuning is what rescues accuracy under the domain shift.
  • The method would degrade gracefully across hyper-parameter settings according to the reported sweep, making it usable without per-dataset tuning.

Reading between the lines

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

  • The same two-stage recipe (LLM-derived concept prompts plus transductive prompt tuning) could plausibly transfer to other domain-shifted detection problems, such as synthetic audio or synthetic text detection, because nothing in the mechanism is face-specific.
  • A fairness test the authors do not report: how much of the 0.098 AUC gain survives when the comparison baselines are also allowed transductive access to the unlabeled test batch, or when test-time tuning runs with shuffled pseudo-labels; without that test, part of the gain may be protocol rather than detector quality.
  • The threshold asymmetry (0.999 for real vs. 0.5 for fake) suggests the selected 'real' samples are trusted far more than the selected 'fake' samples, so a natural extension is to test whether the tuning gain is driven mostly by the high-confidence real subset.
  • One could test the domain-shift story directly: on a test set drawn from the same distribution as training, the test-time tuning gain should shrink toward zero, and a large residual gain there would point to a mechanism other than shift correction.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper proposes a deepfake facial image detection method built on a frozen CLIP backbone with only 768 trainable parameters. It combines knowledge-guided prompt learning (KGP), which uses concepts retrieved from a large language model (e.g., GPT-4) to construct forgery-related prompts, with test-time prompt tuning (TTP), which selects high-confidence unlabeled test samples and tunes the prompts using pseudo-labels generated by the model itself. On the DeepFakeFaceForensics dataset, the method reports an average AUC of 0.911 and an overall accuracy of 0.774, outperforming several inductive baselines including GLFF and UnivFD. The central claim is that this represents a notable improvement over state-of-the-art methods for universal deepfake detection.

Significance. If the reported results hold under a fair protocol, the paper would make a useful contribution: it demonstrates that external knowledge from LLMs can be injected into prompt learning for deepfake detection with very few trainable parameters, and it provides a clean ablation showing that the knowledge-guided prompts alone (without test-time adaptation) already improve over a fixed-CLIP baseline. The idea of using LLM-retrieved semantic concepts to guide prompts is novel in this application and could transfer to other forensic tasks. However, the headline superiority claim is currently confounded by the transductive nature of the test-time tuning stage, which uses the unlabeled test set before evaluation while all baselines are inductive. The significance of the work therefore hinges on whether the evaluation protocol can be made fair and whether the gains from TTP can be attributed to genuine domain-shift correction rather than self-training artifacts.

major comments (3)
  1. [Section II-B, Eqs. (9)-(13); Table III] The comparison with state-of-the-art methods is not apples-to-apples because the proposed method is transductive while every baseline in Tables I and II is inductive. Test-time prompt tuning (TTP) uses the unlabeled test set to select high-confidence samples via Eqs. (9)-(11), generates pseudo-labels, and updates the prompts with the loss in Eq. (13) before the reported AUC/OA are computed. All baselines never touch the test data. This protocol asymmetry is load-bearing: the ablation in Table III shows that removing TTP reduces average OA from 0.774 to 0.598, which is below UnivFD's 0.677, while AUC drops from 0.911 to 0.903. Thus the claimed superiority (e.g., +0.098 AUC over GLFF) is largely attributable to test-set access, not to the knowledge-guided prompt learning alone. The authors should either compare against test-time adaptation or transductive methods under the same protocol, or report the inductive variant as the headline result and treat the TTP gain as a separate contribution.
  2. [Section II-B, Eqs. (9)-(10)] The test-time tuning loop is self-referential: the same model that generates pseudo-labels is then tuned on those labels. With Tfake = 0.5 and TopK = 128, the selected 'fake' set consists of any sample the model currently considers more likely fake than real, so the optimization can reinforce the model's own biases rather than correcting genuine domain shift. The paper does not analyze the accuracy of the pseudo-labels or compare against an oracle variant that uses ground-truth labels during test-time tuning. Without such an analysis, it is unclear whether the TTP improvement reflects a real adaptation to the target domain or merely confirmation bias. Please provide pseudo-label accuracy statistics and an oracle-label experiment to support the claim that TTP alleviates domain shift.
  3. [Section II-B and Section III-A] It is not specified whether the reported AUC and OA are computed on the entire test subset or only on the selected high-confidence samples used for TTP. The selection procedure in Eqs. (9)-(11) with TopK=128 suggests that only a small subset may be used for tuning, but the evaluation could be on all test data in each 'Testing Subset' of Table I. If the metrics are restricted to the selected samples, the results are not comparable to baselines, which are evaluated on the full test distribution. The paper should explicitly state the evaluation set for each metric and, if necessary, report results on the full test set to make the comparison fair.
minor comments (5)
  1. [Section II-B] The sentence 'we calculate the similarity scores and sort them in descending order' is ambiguous: it is unclear whether the sorting is done globally over all test samples or per class, and whether the selection is performed independently for each post-processing subset or over the entire test set. This affects the interpretation of TopK and the reproducibility of the method.
  2. [Section II-A, Eqs. (1)-(2)] The notation for prompts is slightly confusing because Preal and P_fake^i both use [P1]...[PN], but the number of learnable prompts is later stated as 1 in the experiments. Please clarify how the shared learnable prompt tokens interact with the multiple LLM-retrieved concepts when N=1, and whether N denotes the number of tokens or the number of prompt vectors.
  3. [Section III-D] There is a typo in the text: 'Comared' should be 'Compared' in the paragraph discussing trainable parameters. Additionally, the hyperparameter plots in Fig. 3 would benefit from error bars or repeated-run statistics, since the OA differences between some settings (e.g., 0.774 vs 0.764) are small and may not be significant.
  4. [Section III-C] The comparison set is somewhat narrow: beyond UnivFD, recent CLIP-based deepfake detectors such as FatFormer and AntiFakePrompt are mentioned in related work but not included in the quantitative comparison. Adding at least one recent prompt-based baseline would strengthen the claim of outperforming state-of-the-art methods.
  5. [Section II-A] The paper does not provide details of the LLM query, the exact set of retrieved concepts, or any filtering/validation process. For reproducibility, please include the query template and the list of concepts used for the fake prompt, or state that all retrieved concepts are used without filtering.

Circularity Check

1 steps flagged · score 4.0 of 10

Test-time prompt tuning is a self-training loop whose reported gain is measured on the same test distribution used for adaptation; the supervised KGP stage is independent.

  1. fitted input called prediction [Section II-B (Eqs. 9-13), Section III-D (Table III)]
    "To address the domain shift, we tune the prompts on testing data, leveraging pseudo labels generated from the model trained with knowledge-prompt learning. ... Without TTP, the performance drops slightly on AUC while it decreases greatly on OA, indicating that TTP can effectively alleviate the large domain shift between training and testing categories with the further adaption of prompts on testing data."

    Eqs. (9)-(10) select Xreal/Xfake using the model's own probabilities delta; Eq. (12) converts those selections into pseudo-labels bY; Eq. (13) optimizes the same prompt with the same cross-entropy loss as Eq. (5) against those pseudo-labels. Table III then reports AUC/OA on the same test set from which the tuning samples were drawn. The TTP gain (e.g., OA 0.598 to 0.774) is therefore partly a self-referential re-fit: the prompt is adapted to the model's own high-confidence hypotheses about the evaluation distribution and then evaluated on that distribution. This does not collapse the whole method, because the KGP stage is supervised and the final metrics use ground-truth labels, but the TTP-based improvement is not independent evidence of domain-shift correction.

full rationale

The core knowledge-guided prompt learning stage is self-contained: it uses ground-truth labels on the LSUN/ProGAN training set (Eq. 5) and an external LLM as the knowledge source, so no circular reduction is present there. The only partially circular component is test-time prompt tuning: pseudo-labels are generated by the model itself, the prompt is updated against those pseudo-labels, and the resulting performance is measured on the same test distribution that supplied the tuning samples. This is a self-training loop rather than a derivation from first principles, and the reported OA improvement in particular is a transductive adaptation effect. The central SOTA claim retains some independent support, since even without TTP the AUC (0.903) exceeds the best inductive baseline GLFF (0.813). No self-citation chain, imported uniqueness theorem, or ansatz-smuggling-by-citation is present. The protocol asymmetry with inductive baselines is a correctness concern, but the TTP step itself is the only place where the result partly reduces to its own inputs.

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

The reported gains depend on a single 768-dimensional learnable prompt token, hand-chosen thresholds, and the assumption that frozen CLIP features plus self-generated pseudo-labels yield reliable test-time adaptation.

free parameters (4)
  • learnable prompt embedding P1 = 768-dim vector, final weights not released
    The single learnable prompt token is optimized by cross-entropy on training data (Section II-A) and further by TTP on pseudo-labeled test data (Section II-B). It is the central fitted quantity.
  • threshold T_real = 0.999
    Hand-chosen in Section III-B for selecting confident real images for TTP; Fig. 3 shows robustness across 0.9 to 0.99999, but it is still a manually set hyperparameter.
  • threshold T_fake = 0.5
    Hand-chosen in Section III-B for selecting fake images for TTP; not justified beyond the robustness figure.
  • TopK = 128
    Maximum number of pseudo-labeled samples per class in TTP; chosen in Section III-B, and the ablation shows performance varies with its value.
assumptions (5)
  • domain assumption Pretrained CLIP ViT-L/14 provides a shared embedding space where textual forgery concepts are discriminative for facial deepfakes.
    The entire method relies on frozen CLIP features without fine-tuning; if this assumption fails, prompt learning cannot help. Invoked in Section III-B.
  • domain assumption The model's own high-confidence predictions constitute a reliable source of pseudo-labels for test-time adaptation.
    Section II-B uses Eqs. 9-11 to select samples; if these pseudo-labels are biased, TTP can degrade performance.
  • domain assumption The distribution shift between LSUN training categories and face test categories is a label/domain shift that prompt tuning can absorb.
    This motivates the test-time tuning in Sections I and II-B.
  • domain assumption GPT-4 responses yield forgery-related concepts that improve prompt semantics.
    Section II-A retrieves concepts such as 'unnatural' and 'blurred' with no validation of which concepts actually matter for detection.
  • domain assumption Published baseline numbers in Table I are comparable despite different training pipelines and evaluation protocols.
    The paper copies AUC numbers from original papers; any discrepancy in preprocessing, data splits, or evaluation details could affect comparability.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Knowledge-Guided Prompt Learning for Deepfake Facial Image Detection." pith.science (2026). https://pith.science/paper/NQA34KMB

@misc{pith2026250100700,
  author       = {Pith},
  title        = {Pith review of: Knowledge-Guided Prompt Learning for Deepfake Facial Image Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NQA34KMB}},
  note         = {Machine review of arXiv:2501.00700}
}
read the original abstract

Recent generative models demonstrate impressive performance on synthesizing photographic images, which makes humans hardly to distinguish them from pristine ones, especially on realistic-looking synthetic facial images. Previous works mostly focus on mining discriminative artifacts from vast amount of visual data. However, they usually lack the exploration of prior knowledge and rarely pay attention to the domain shift between training categories (e.g., natural and indoor objects) and testing ones (e.g., fine-grained human facial images), resulting in unsatisfactory detection performance. To address these issues, we propose a novel knowledge-guided prompt learning method for deepfake facial image detection. Specifically, we retrieve forgery-related prompts from large language models as expert knowledge to guide the optimization of learnable prompts. Besides, we elaborate test-time prompt tuning to alleviate the domain shift, achieving significant performance improvement and boosting the application in real-world scenarios. Extensive experiments on DeepFakeFaceForensics dataset show that our proposed approach notably outperforms state-of-the-art methods.

Figures

Figures reproduced from arXiv: 2501.00700 by the authors.

Figure 1
Figure 1. Existing approaches mostly focus on forgery artifacts extraction. [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. The overall framework, which consists of knowledge-guided prompt learning and test-time prompt tuning. The former elicit prior knowledge from a [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Detection performance with different value of each hyper-parameter. [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

36 extracted references · 24 canonical work pages

  1. [1]

    Generative adversarial networks,

    I. Goodfellow, J. Pouget-Abadie, M. Mirza, B. Xu, D. Warde-Farley, S. Ozair, A. Courville, and Y . Bengio, “Generative adversarial networks,” Communications of the ACM , vol. 63, no. 11, pp. 139–144, 2020

    2020

  2. [2]

    Denoising diffusion probabilistic models,

    J. Ho, A. Jain, and P. Abbeel, “Denoising diffusion probabilistic models,” in Neural Information Processing Systems , vol. 33, 2020, pp. 6840– 6851

    work page 2020

  3. [3]

    Cnn- generated images are surprisingly easy to spot... for now,

    S.-Y . Wang, O. Wang, R. Zhang, A. Owens, and A. A. Efros, “Cnn- generated images are surprisingly easy to spot... for now,” in IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2020, pp. 8695–8704

    work page 2020

  4. [4]

    Attributing fake images to gans: Learning and analyzing gan fingerprints,

    N. Yu, L. S. Davis, and M. Fritz, “Attributing fake images to gans: Learning and analyzing gan fingerprints,” in IEEE/CVF International Conference on Computer Vision , 2019, pp. 7556–7566

    work page 2019

  5. [5]

    Detection of gan-generated fake images over social networks,

    F. Marra, D. Gragnaniello, D. Cozzolino, and L. Verdoliva, “Detection of gan-generated fake images over social networks,” inIEEE Conference on Multimedia Information Processing and Retrieval , 2018, pp. 384–389

    work page 2018

  6. [6]

    What makes fake images detectable? understanding properties that generalize,

    L. Chai, D. Bau, S.-N. Lim, and P. Isola, “What makes fake images detectable? understanding properties that generalize,” in European Con- ference on Computer Vision , 2020, pp. 103–120

    work page 2020

  7. [7]

    Finding facial forgery artifacts with parts-based detectors,

    S. Schwarcz and R. Chellappa, “Finding facial forgery artifacts with parts-based detectors,” in IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2021, pp. 933–942

    work page 2021

  8. [8]

    Multi- attentional deepfake detection,

    H. Zhao, W. Zhou, D. Chen, T. Wei, W. Zhang, and N. Yu, “Multi- attentional deepfake detection,” in IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2021, pp. 2185–2194

    work page 2021

Show all 36 references
  1. [9]

    Patch-dfd: Patch-based end-to-end deepfake discriminator,

    M. Yu, S. Ju, J. Zhang, S. Li, J. Lei, and X. Li, “Patch-dfd: Patch-based end-to-end deepfake discriminator,” Neurocomputing, vol. 501, pp. 583– 595, 2022

    2022

  2. [10]

    Detecting gan- generated images by orthogonal training of multiple cnns,

    S. Mandelli, N. Bonettini, P. Bestagini, and S. Tubaro, “Detecting gan- generated images by orthogonal training of multiple cnns,” in IEEE International Conference on Image Processing , 2022, pp. 3091–3095

    2022

  3. [11]

    Leveraging frequency analysis for deep fake image recog- nition,

    J. Frank, T. Eisenhofer, L. Sch ¨onherr, A. Fischer, D. Kolossa, and T. Holz, “Leveraging frequency analysis for deep fake image recog- nition,” in International Conference on Machine Learning , 2020, pp. 3247–3258

    2020

  4. [12]

    Thinking in frequency: Face forgery detection by mining frequency-aware clues,

    Y . Qian, G. Yin, L. Sheng, Z. Chen, and J. Shao, “Thinking in frequency: Face forgery detection by mining frequency-aware clues,” in European Conference on Computer Vision , 2020, pp. 86–103

    2020

  5. [13]

    Frequency masking for universal deepfake detection,

    C. T. Doloriel and N.-M. Cheung, “Frequency masking for universal deepfake detection,” in IEEE International Conference on Acoustics, Speech and Signal Processing , 2024, pp. 13 466–13 470

    2024

  6. [14]

    Glff: Global and local feature fusion for ai-synthesized image detection,

    Y . Ju, S. Jia, J. Cai, H. Guan, and S. Lyu, “Glff: Global and local feature fusion for ai-synthesized image detection,” IEEE Transactions on Multimedia , vol. 26, pp. 4073–4085, 2023

    2023

  7. [15]

    Frequency-aware attentional feature fusion for deepfake detection,

    C. Tian, Z. Luo, G. Shi, and S. Li, “Frequency-aware attentional feature fusion for deepfake detection,” in IEEE International Conference on Acoustics, Speech and Signal Processing , 2023, pp. 1–5

    2023

  8. [16]

    Towards universal fake image detectors that generalize across generative models,

    U. Ojha, Y . Li, and Y . J. Lee, “Towards universal fake image detectors that generalize across generative models,” in IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2023, pp. 24 480–24 489

    2023

  9. [17]

    Learning transferable visual models from natural language supervision,

    A. Radford, J. W. Kim, C. Hallacy, A. Ramesh, G. Goh, S. Agarwal, G. Sastry, A. Askell, P. Mishkin, J. Clark et al. , “Learning transferable visual models from natural language supervision,” in International Conference on Machine Learning , 2021, pp. 8748–8763

    2021

  10. [18]

    Forgery-aware adaptive transformer for generalizable synthetic image detection,

    H. Liu, Z. Tan, C. Tan, Y . Wei, J. Wang, and Y . Zhao, “Forgery-aware adaptive transformer for generalizable synthetic image detection,” in IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2024, pp. 10 770–10 780

    2024

  11. [19]

    Anomalyclip: Object-agnostic prompt learning for zero-shot anomaly detection,

    Q. Zhou, G. Pang, Y . Tian, S. He, and J. Chen, “Anomalyclip: Object-agnostic prompt learning for zero-shot anomaly detection,” arXiv preprint arXiv:2310.18961, 2023

    2023

  12. [20]

    Promptad: Learning prompts with only normal samples for few-shot anomaly detection,

    X. Li, Z. Zhang, X. Tan, C. Chen, Y . Qu, Y . Xie, and L. Ma, “Promptad: Learning prompts with only normal samples for few-shot anomaly detection,” in IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 16 838–16 848

    2024

  13. [21]

    Gpt-4 technical report,

    J. Achiam, S. Adler, S. Agarwal, L. Ahmad, I. Akkaya, F. L. Aleman, D. Almeida, J. Altenschmidt, S. Altman, S. Anadkat et al. , “Gpt-4 technical report,” arXiv preprint arXiv:2303.08774 , 2023

    2023 arXiv

  14. [22]

    Conditional prompt learning for vision-language models,

    K. Zhou, J. Yang, C. C. Loy, and Z. Liu, “Conditional prompt learning for vision-language models,” in IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2022, pp. 16 816–16 825

    2022

  15. [23]

    Maple: Multi-modal prompt learning,

    M. U. Khattak, H. Rasheed, M. Maaz, S. Khan, and F. S. Khan, “Maple: Multi-modal prompt learning,” in IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2023, pp. 19 113–19 122

    2023

  16. [24]

    Conditioned prompt-optimization for continual deepfake detection,

    F. Laiti, B. Liberatori, T. De Min, and E. Ricci, “Conditioned prompt-optimization for continual deepfake detection,” arXiv preprint arXiv:2407.21554, 2024

    2024 arXiv

  17. [25]

    Antifakeprompt: Prompt- tuned vision-language models are fake image detectors,

    Y .-M. Chang, C. Yeh, W.-C. Chiu, and N. Yu, “Antifakeprompt: Prompt- tuned vision-language models are fake image detectors,” arXiv preprint arXiv:2310.17419, 2023

    2023 arXiv

  18. [26]

    Are gan generated images easy to detect? a critical analysis of the state- of-the-art,

    D. Gragnaniello, D. Cozzolino, F. Marra, G. Poggi, and L. Verdoliva, “Are gan generated images easy to detect? a critical analysis of the state- of-the-art,” in IEEE International Conference on Multimedia and Expo , 2021, pp. 1–6

    2021

  19. [27]

    Beyond the spectrum: Detecting deepfakes via re-synthesis,

    Y . He, N. Yu, M. Keuper, and M. Fritz, “Beyond the spectrum: Detecting deepfakes via re-synthesis,” arXiv preprint arXiv:2105.14376 , 2021

    2021 arXiv

  20. [28]

    Fusing global and local features for generalized ai-synthesized image detection,

    Y . Ju, S. Jia, L. Ke, H. Xue, K. Nagano, and S. Lyu, “Fusing global and local features for generalized ai-synthesized image detection,” in IEEE International Conference on Image Processing , 2022, pp. 3465–3469

    2022

  21. [29]

    Lsun: Construction of a large-scale image dataset using deep learning with humans in the loop,

    F. Yu, A. Seff, Y . Zhang, S. Song, T. Funkhouser, and J. Xiao, “Lsun: Construction of a large-scale image dataset using deep learning with humans in the loop,” arXiv preprint arXiv:1506.03365 , 2015

    2015 arXiv

  22. [30]

    Progressive growing of gans for improved quality, stability, and variation,

    T. Karras, “Progressive growing of gans for improved quality, stability, and variation,” arXiv preprint arXiv:1710.10196 , 2017

    2017 arXiv

  23. [31]

    An image is worth 16x16 words: Transformers for image recognition at scale,

    A. Dosovitskiy, “An image is worth 16x16 words: Transformers for image recognition at scale,” arXiv preprint arXiv:2010.11929 , 2020

    2010 arXiv

  24. [32]

    Adam: A method for stochastic optimization,

    D. P. Kingma, “Adam: A method for stochastic optimization,” arXiv preprint arXiv:1412.6980, 2014

    2014 arXiv

  25. [33]

    Towards evaluating the robustness of neural networks,

    N. Carlini and D. Wagner, “Towards evaluating the robustness of neural networks,” in IEEE Symposium on Security and Privacy , 2017, pp. 39– 57

    2017

  26. [34]

    Ganprintr: Improved fakes and evaluation of the state of the art in face manipulation detection,

    J. C. Neves, R. Tolosana, R. Vera-Rodriguez, V . Lopes, H. Proenc ¸a, and J. Fierrez, “Ganprintr: Improved fakes and evaluation of the state of the art in face manipulation detection,” IEEE Journal of Selected Topics in Signal Processing, vol. 14, no. 5, pp. 1038–1048, 2020

    2020

  27. [35]

    Deep residual learning for image recognition,

    K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” in IEEE Conference on Computer Vision and Pattern Recognition, 2016, pp. 770–778

    2016

  28. [36]

    Yfcc100m: The new data in multimedia research,

    B. Thomee, D. A. Shamma, G. Friedland, B. Elizalde, K. Ni, D. Poland, D. Borth, and L.-J. Li, “Yfcc100m: The new data in multimedia research,” Communications of the ACM , vol. 59, no. 2, pp. 64–73, 2016

    2016

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.