Pith. sign in

REVIEW 4 major objections 5 minor 38 references

FairForensics: Seeing Expressions and Parsing Demographics via Vision-Language Modeling for Generalizable Fair Deepfake Detection

T0 review · 4 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read FairForensics pairs expression cues with demographic language prompts to make deepfake detection both more generalizable and fairer across demographic groups.

desk verdict Real cross-dataset gains and a useful balanced fairness benchmark, but the VFA loss and IAPP mechanism are misdescribed as written, and the 308-token text encoder detail is unresolved. read the letter →

arxiv 2608.01661 v1 pith:YOLFZ4RQ submitted 2026-08-03 cs.CV

Yaning Zhang , Jiao Wu , Zan Gao , Linlin Shen This is my paper · ORCID
classification cs.CV
keywords fairdeepfakedetectionvision-languagemodeldemographicfairnessfacialexpressionforgeryidentity-awarepatchperturbationprototypecross-datasetgeneralizationbalancedbenchmark
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 detection can be made both more accurate on unseen forgeries and fairer across demographic groups by combining two signals that prior detectors mostly ignore: high-level facial-expression differences between real and fake faces, and language prompts that name a face's demographic attributes. It introduces FairForensics, whose central claim is that expression-guided forgery features, identity-suppressing patch perturbations, demographic-aware text supervision, and prototype-level fairness regularization together produce representations that are discriminative for forgery yet invariant to identity and demographics. To support this, the paper builds a demographically balanced benchmark from GenFace with eight intersectional subgroups (sex, age, and race) and evaluates under balanced and imbalanced training as well as cross-dataset transfer to FF++, Celeb-DF, DFDC, and DF-1.0. If correct, the work matters because it offers a concrete recipe for stopping detectors from latching onto who a face is rather than whether it was manipulated.

What carries the argument

The machinery is a multi-module training pipeline. The expression encoder (EE) takes the 8-dimensional embedding from a frozen POSTER expression estimator and adapts it through a trainable adapter and forgery projector, turning expression vectors into expression-guided forgery features supervised by their own real/fake loss. The expression-perceptual visual encoder (EPVE) is a vision transformer whose blocks contain an expression injector (EI): the class token queries the expression features in parallel attention, so global appearance tokens learn to absorb expression forgery cues. After the first block, an identity-aware patch perturbation module (IAPP) scores each patch token for identity sensitivity, then during training adds Gaussian noise scaled by that score and drops a subset of the top-k identity-responsive patches, forcing the model to rely on non-identity regions. The demographic-guided language encoder (DGLE) builds four prompts (authenticity, sex, age, race), tokenizes each to 77 tokens, concatenates them into a 308-token sequence, and encodes it with a text transformer to obtain population-aware language embeddings aligned to visual features through a contrastive loss. Finally, the population-aware prototype fairness objective (PPF) computes class-conditional subgroup prototypes and applies a margin loss separating real and fake prototypes across groups plus an alignment loss pulling same-class prototypes together; these are combined with visual forgery, expression forgery, vision-vision contrastive, and language discrimination losses in a single objective.

What would settle it

Open the released code and feed a 308-token input to the text encoder: if the encoder is the standard CLIP text tower with 77-token positional embeddings, it will either error, truncate, or silently repeat positions, and inspecting whether the positional embedding table was interpolated or replaced would settle whether the demographic language branch works as described. A second check is to train the full model with the demographic language branch removed and compare cross-dataset AUC and fairness on Celeb-DF; the paper's own ablations predict a large drop, so observing no drop would cast doubt on the mechanism.

Watch

Extended reading notes

Core claim

The paper claims that deepfake detection can be simultaneously generalized and made fair by treating facial expression as a structured forgery signal and demographic attributes as language-level supervision. FairForensics builds on the observation that real and fake faces differ in their expression-embedding distributions, and uses that difference to inject expression-aware forgery cues into a vision transformer while an identity-aware patch perturbation suppresses person-specific shortcuts. On the demographic side, text prompts describing authenticity, sex, age, and race are encoded together and aligned with visual features, and a prototype-level objective pulls same-class demographic subgroups together while pushing real and fake prototypes apart. On its balanced GenFace benchmark and cross-dataset tests, the paper reports AUC of 63.04% on FF++, 62.35% on Celeb-DF, 59.88% on DFDC, and 62.75% on DF-1.0, with lower false-positive-rate and equalized-odds gaps than the compared detectors, and it states that this is state of the art in both generalization and fairness.

Load-bearing premise

The load-bearing premise of the demographic language branch is that a CLIP-style text encoder can ingest the concatenated 308-token prompt, yet the paper never specifies how the encoder's 77-token positional embeddings are extended, so if that step is unstated the branch cannot operate as described.

Editorial extensions

If this is right

  • Balanced demographic training data alone lowers fairness gaps and improves cross-dataset AUC relative to imbalanced training for the same detector.
  • Adding sex, age, and race text prompts on top of authenticity prompts improves both detection AUC and demographic fairness on unseen datasets.
  • Perturbing identity-sensitive patches during training reduces identity-shortcut reliance and improves fairness on cross-dataset tests with only a small parameter increase.
  • Removing either the prototype margin loss or the subgroup alignment loss degrades cross-dataset AUC and fairness, so both components are needed.
  • The model is more robust to unseen image distortions such as Gaussian noise, blur, and compression than the compared detectors.

Reading between the lines

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

  • If the expression-distribution signal is as general as the paper argues, any frozen expression estimator whose real/fake embeddings are well separated could serve as the forgery prior, so a cheap test is to swap POSTER for another estimator and measure cross-dataset AUC.
  • The paper's Table IV shows that inaccurate race prompts hurt fairness, implying that demographic annotation noise is a sensitivity point; a natural extension is evaluating the framework under noisy or missing demographic labels.
  • The 308-token text-encoder question is testable in the released code: if the standard 77-token CLIP position embeddings are not extended, the demographic language branch behaves differently from the description, and the fairness gains attributed to it would need re-attribution.
  • The benchmark's demographic bins (0-29 versus 30 and over, white versus non-white, male versus female) are coarse, and the paper itself notes residual bias under complex intersections, so extending the prototype and language machinery to finer-grained or continuous attributes is a natural next step.
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

4 major / 5 minor

Summary. The paper presents FairForensics, a vision-language framework for fair and generalizable deepfake detection on a demographically balanced benchmark constructed from GenFace. It combines an expression encoder built on POSTER, an expression-perceptual visual encoder with an identity-aware patch perturbation (IAPP) module, a demographic-guided language encoder (DGLE) that builds text prompts over authenticity, sex, age, and race, and a population-aware prototype fairness (PPF) objective. The authors report state-of-the-art cross-dataset AUC and fairness gaps on FF++, Celeb-DF, DFDC, and DF-1.0, with extensive ablations supporting the role of each component.

Significance. If validated, the balanced benchmark itself would be a useful resource for the fairness community, and the cross-dataset results suggest that demographic-aware language prompts and balanced training distributions can improve fairness without sacrificing detection performance. The paper ships no code or checkpoints at submission, and several named mechanisms conflict with the equations provided, so the current evidence for the specific design choices is incomplete. The ablation suite is broad (Tables IV-VIII) and the use of genuinely held-out public datasets for the headline claim is a real strength.

major comments (4)
  1. [III-D, Eq. (7)-(8)] The VFA loss is claimed to 'amplify forgery features' and to 'pull positive pairs together while pushing negative ones away,' but the positive pair for each anchor is the feature of the same image itself. The one-hot label y_pa in Eq. (8) sets y^uu_pa = 1 and y^uv_pa = 0 for u ≠ v, so Eq. (7) is an instance-discrimination softmax over the batch, not a forgery-amplification objective. No authenticity labels enter the loss, so real-real and real-fake pairs are pushed apart with equal strength. Table V ascribes a 3.79% AUC gain on DFDC to L_vfa, but a generic feature-spreading effect could produce that gain; the claimed expression-aware amplification mechanism is not realized by the published equations. The authors should either reformulate the loss with genuinely forgery-relevant positives (e.g., same-class or expression-matched pairs) or remove the unsupported mechanistic claim.
  2. [III-C, DGLE and text encoder] T_emb is defined as being in R^{b x 308 x s} and is fed to the text encoder with a positional embedding P_t in R^{b x 308 x s}. The paper cites CLIP [14] as the text encoder and tokenizer; standard CLIP text transformers have a fixed 77-token context length with learned positional embeddings of size 77. No modification is described for handling 308 tokens, so as written the demographic language branch cannot be executed. Please specify the architecture change (e.g., extended positional embeddings, per-sentence encoding followed by pooling, or a different text encoder) and update the parameter counts in Table III accordingly.
  3. [III-B, Eq. (4) and IAPP] The identity sensitivity scores S = sigmoid(f_id(I_pat_2)) are produced by a two-layer MLP, but no identity labels, identity classification loss, or identity-contrastive objective is defined anywhere in the paper. Without any supervision linking f_id's output to identity, there is no reason the top-k selected patches correspond to identity-responsive regions; the scores could be driven by forgery cues or arbitrary image statistics. The claim that IAPP 'suppresses identity-related shortcut learning' is therefore unsupported. Add an identity-aware training signal or provide direct evidence (e.g., correlation with identity classification accuracy on the selected patches).
  4. [IV-A, benchmark and baselines] The balanced GenFace benchmark is a re-split of the authors' own dataset with FairFace annotations, and the baseline methods are re-implemented by the authors using 'default configurations' with no released code. Since the protocol is new and baselines are not official checkpoints, the within-dataset comparisons in Table II and the SOTA claim in the abstract cannot be independently verified without releasing the evaluation code, the split indices, and the trained models. At minimum, release the benchmark construction code and the exact train/validation/test identity-disjoint splits; ideally provide official-baseline results on the same splits.
minor comments (5)
  1. [References / Section III-B] Reference [24] is cited for POSTER in Section III-B but [24] is the ETH-XGaze gaze-estimation paper; the correct POSTER citation is [2]. The reference list also lacks the actual POSTER entry under [24].
  2. [IV-C, Table IV discussion] The text says that inaccurate demographic prompts make fairness scores 'improve noticeably,' while the table shows F_FPR and F_MEO increase (worsen) for both MFCLIP and FairForensics (e.g., Ours w/ASAR* has F_FPR 10.45 vs Ours w/ASAR† with 16.24 on FF++). Please reword to match the reported numbers.
  3. [Table IV] The row label 'Ours w/ASA' appears to be a typo for 'Ours w/ASG', given the column headings A, S, G, and R.
  4. [Table III] There are punctuation typos that impede reading: 'Params 421.587' should likely be '421.59', and the CViT row shows '49,89' with a comma instead of a decimal point.
  5. [III-D, DSAD description] The text 'senmatic signals' contains a typo for 'semantic signals'; also the notation I^r_vA and I^f_vA in the L_vlc explanation is not defined in the notation list.

Circularity Check

1 steps flagged · score 4.0 of 10

Cross-dataset evaluation on public benchmarks keeps the central generalization/fairness claim independent, but the VFA loss as written is a self-positive instance-contrastive objective renamed as expression-aware forgery amplification, and the benchmark reuses the authors' own GenFace data.

  1. renaming known result [Section III-D, Eqs. (7)-(8)]
    "for the vision-vision feature pairs {(I^u_v, I^u_v)}^b_{u=1}, we calculate the vision-to-vision cosine similarity vector ... The one-hot label y_pa of the u-th pair is denoted as y^u_pa = {y^{uv}_pa}^b_{v=1}, y^{uu}_pa = 1, y^{uv,u≠v}_pa = 0. To pull positive pairs together while pushing negative ones away, the VFA loss is denoted as follows: L_vfa = ..."

    In Eq. (7), the softmax is over cosine similarities between anchor I^u_v and every I^v_v in the batch; Eq. (8) assigns the positive label only to u=v. Thus each sample's own feature is its sole positive, and no authenticity label enters the loss. The objective is exactly an instance-level InfoNCE/self-contrastive loss: it spreads all samples apart, repelling real-real pairs as strongly as real-fake pairs. The paper's claim that this loss amplifies expression-aware forgery features is therefore not a consequence of the equation; it is a renamed standard instance-discrimination objective. Consequently, the 3.79% DFDC AUC drop in Table V when removing L_vfa is attributable, by the equation itself, to generic feature dispersion rather than to forgery-specific amplification.

full rationale

The paper's central SOTA claim is not reduced to a fitted parameter: cross-dataset generalization and fairness are evaluated on held-out public datasets (FF++, Celeb-DF, DFDC, DF-1.0), and the main components (expression encoder, EPVE, DGLE, PPF) are trained with explicit loss terms and tested by ablations. The benchmark is a re-split of the authors' own GenFace dataset and several baselines are re-implemented by the same authors, which limits external verification but is not circular derivation. The self-citations (MFCLIP [1], GenFace [17]) are used as baselines/data sources, not as a uniqueness theorem or a load-bearing premise that forces the result. The one equation-level concern is the VFA loss: as written it is a self-positive instance-contrastive loss with no authenticity labels, so calling it 'visual forgery amplification' renames a known generic contrastive mechanism rather than realizing expression-aware forgery supervision. The 308-token DGLE sequence also exceeds CLIP's 77-token positional embedding without explanation, but that is an implementation gap, not circularity. On balance, the central claim has independent content, but the VFA component and the self-citation-heavy benchmark warrant a moderate score.

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

The method depends on several unstated or untested premises: the text encoder can handle 308 tokens, POSTER expression embeddings carry transferable forgery signal, the identity scoring head can learn identity sensitivity without identity labels, and FairFace annotations are accurate. There are no invented physical entities. The main tuned hyperparameters (r, margin, lambda, sigma^2) are fixed without sensitivity analysis.

free parameters (4)
  • r (selection ratio) = 0.5
    Proportion of top identity-sensitive patches perturbed in IAPP; set without sensitivity analysis.
  • margin (mar) = 0.5
    Minimum distance between real and fake prototypes in PPF margin loss.
  • loss weight lambda = 1.0
    Weight of PPF loss in the total objective.
  • Gaussian noise std sigma^2 = 0.05
    Standard deviation of perturbation in IAPP.
assumptions (4)
  • ad hoc to paper The standard CLIP text encoder can process 308-token sequences
    The demographic language encoder concatenates four 77-token sentences into a 308-token input, but no modification of CLIP's 77-positional-embedding text encoder is described.
  • domain assumption Expression embeddings from frozen POSTER contain forgery-relevant signals that transfer across datasets
    The FID-based motivation is established only on GenFace; transfer to FF++, Celeb-DF, DFDC, and DF-1.0 is assumed.
  • ad hoc to paper Identity sensitivity can be learned without identity supervision
    The IAPP scoring head has no identity labels or auxiliary loss in the paper; the identity-aware property is asserted.
  • domain assumption FairFace demographic annotations are accurate enough
    All demographic labels and fairness metrics depend on automatic FairFace annotation; annotation errors would shift the metrics.

how reviews work

0 comments
Cite this review

Pith. "Pith review of FairForensics: Seeing Expressions and Parsing Demographics via Vision-Language Modeling for Generalizable Fair Deepfake Detection." pith.science (2026). https://pith.science/paper/YOLFZ4RQ

@misc{pith2026260801661,
  author       = {Pith},
  title        = {Pith review of: FairForensics: Seeing Expressions and Parsing Demographics via Vision-Language Modeling for Generalizable Fair Deepfake Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YOLFZ4RQ}},
  note         = {Machine review of arXiv:2608.01661}
}
read the original abstract

The challenge of fair deepfake detection (FDD) has attracted increasing attention. Existing fairness-enhanced detectors often suffer from suboptimal generalization to unseen manipulations and fairness across demographic groups. They are typically developed and evaluated on demographically imbalanced distributions, resulting in biased predictions toward minority groups. In this paper, we construct a novel demographically balanced FDD benchmark to train and evaluate the fairness of detectors under both balanced and imbalanced population scenarios. Additionally, we introduce a novel expression and demographic perceptual vision-language model, termed FairForensics, for generalizable fair deepfake detection. FairForensics conducts face forgery generalization enhancement and demographic-aware fairness regularization. During face forgery generalization enhancement, built upon the novel observation of significant distribution differences between pristine and forged expression vectors, we design an expression encoder to capture high-level expression-guided forgery patterns, and an expression-perceptual visual encoder that integrates global appearance and expression forgery features while mitigating identity bias using an identity-aware patch perturbation module. Under demographic-aware fairness regularization, we propose a demographic-guided language encoder to extract population-aware global language embeddings, which boosts the decoupling of forgery features from demographic information via vision-language alignment. We devise a population-aware prototype fairness objective to enforce both inter-class separability and intra-class alignment across demographic subgroups. Extensive experiments on our balanced demographic benchmark show that our method achieves the state-of-the-art in terms of generalization and fairness.

Figures

Figures reproduced from arXiv: 2608.01661 by the authors.

Figure 1
Figure 1. (a) t-SNE visualization of various intersectional demo [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. (a) Existing fairness-enhanced models mainly rely [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. The workflow of our FairForensics model. After obtaining multiple patches and demographic texts of the input face [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: The pipeline of IAPP. A photo of a fake face A photo of a male face A photo of a face aged 30 and over A photo of a face with a white skin tone Real Fake Female Male 0-29 30+ Non-White White Demographic-Aware Text Prompts Authenticity Forgery Sex Age Race Female Sex Ag…
Figure 5
Figure 5. Figure 5: The architecture of the demographic text generator. [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: Resilience to unseen image distortions. visual embeddings with these language features encourages the model to distinguish real and fake faces under different demographic contexts. For example, for real visual features I r vA and fake ones I f vA from the same demograp…
Figure 7
Figure 7. Figure 7: t-SNE visualization of expression embeddings extracted by POSTER pretrained using AffectNet8 and SSFER pretrained [PITH_FULL_IMAGE:figures/full_fig_p009_7.png]
Figure 8
Figure 8. Figure 8: Heatmap visualizations of fairness-enhanced detectors [PITH_FULL_IMAGE:figures/full_fig_p011_8.png]
Figure 9
Figure 9. Figure 9: The detection performance of various detectors trained [PITH_FULL_IMAGE:figures/full_fig_p012_9.png]
Figure 10
Figure 10. Figure 10: Visualization of prototype distribution and intra-/cross-group distance heatmaps of prototypes generated by our model [PITH_FULL_IMAGE:figures/full_fig_p013_10.png]

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

38 extracted references · 18 canonical work pages

  1. [14]

    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. Clarket al., “Learning transferable visual models from natural language supervision,” inProceedings of the International Conference on Machine Learning (ICML), 2021, pp. 8748–8763

    2021

  2. [1]

    Mfclip: Multi-modal fine-grained clip for generalizable diffusion face forgery detection,

    Y . Zhang, T. Wang, Z. Yu, Z. Gao, L. Shen, and S. Chen, “Mfclip: Multi-modal fine-grained clip for generalizable diffusion face forgery detection,”IEEE Transactions on Information Forensics and Security, vol. 20, pp. 5888–5903, 2025

    2025

  3. [2]

    Poster: A pyramid cross-fusion transformer network for facial expression recognition,

    C. Zheng, M. Mendieta, and C. Chen, “Poster: A pyramid cross-fusion transformer network for facial expression recognition,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 3146–3155

    work page 2023

  4. [3]

    Generative adversarial nets,

    I. Goodfellow, J. Pouget-Abadie, M. Mirza, B. Xu, D. Warde-Farley, S. Ozair, A. Courville, and Y . Bengio, “Generative adversarial nets,” in Advances in Neural Information Processing Systems, 2014, pp. 2672– 2680

    2014

  5. [4]

    DiffFace: Diffusion-based face swapping with facial guidance,

    K. Kim, Y . Kim, S. Cho, J. Seo, J. Nam, K. Lee, S. Kim, and K. Lee, “DiffFace: Diffusion-based face swapping with facial guidance,”arXiv preprint arXiv:2212.13344, 2022

    arXiv 2022

  6. [5]

    Xception: Deep learning with depthwise separable convolu- tions,

    F. Chollet, “Xception: Deep learning with depthwise separable convolu- tions,” inProceedings of the IEEE conference on computer vision and pattern recognition, 2017, pp. 1251–1258

    2017

  7. [6]

    Very deep convolutional networks for large-scale image recognition,

    K. Simonyan and A. Zisserman, “Very deep convolutional networks for large-scale image recognition,” inInternational Conference on Learning Representations (ICLR), 2015, pp. 1–14

    work page 2015

  8. [7]

    Deepfake video detection using convolutional vision transformer,

    D. Wodajo and S. Atnafu, “Deepfake video detection using convolutional vision transformer,” 2021, arXiv preprint arXiv:2102.11126

    arXiv 2021

Show all 38 references
  1. [8]

    Gbdf: Gender balanced deepfake dataset towards fair deepfake detection,

    A. V . Nadimpalli and A. Rattani, “Gbdf: Gender balanced deepfake dataset towards fair deepfake detection,” inInternational Conference on Pattern Recognition. Springer, 2022, pp. 320–337

    2022

  2. [9]

    Ai-face: A million- scale demographically annotated ai-generated face dataset and fairness benchmark,

    L. Lin, S. Santosh, M. Wu, X. Wang, and S. Hu, “Ai-face: A million- scale demographically annotated ai-generated face dataset and fairness benchmark,” inProceedings of the Computer Vision and Pattern Recog- nition Conference, 2025, pp. 3503–3515

    2025

  3. [10]

    Crossdf: improving cross-domain deepfake detection with deep infor- mation decomposition,

    S. Yang, H. Guo, S. Hu, B. Zhu, Y . Fu, S. Lyu, X. Wu, and X. Wang, “Crossdf: improving cross-domain deepfake detection with deep infor- mation decomposition,”Frontiers in Big Data, vol. 8, p. 1669488, 2025

    2025

  4. [11]

    Preserving fairness generalization in deepfake detection,

    L. Lin, X. He, Y . Ju, X. Wang, F. Ding, and S. Hu, “Preserving fairness generalization in deepfake detection,” inProceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2024, pp. 16 815–16 825

    2024

  5. [12]

    Decoupling bias, aligning distributions: Synergistic fairness optimization for deepfake detection,

    F. Ding, W. Yi, Y . Zhou, X. He, H. Rao, and S. Hu, “Decoupling bias, aligning distributions: Synergistic fairness optimization for deepfake detection,” inProceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2026, pp. 32 704–32 713

    2026

  6. [13]

    Redundant semantic environment filling via misleading-learning for fair deepfake detection,

    X. He, Y . Zhou, S. Hu, B. Li, J. Huang, and F. Ding, “Redundant semantic environment filling via misleading-learning for fair deepfake detection,”arXiv preprint arXiv:2405.15173, 2024

    2024

  7. [15]

    Implicit identity leakage: The stumbling block to improving deepfake detection generalization,

    S. Dong, J. Wang, R. Ji, J. Liang, H. Fan, and Z. Ge, “Implicit identity leakage: The stumbling block to improving deepfake detection generalization,” in2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2023, pp. 3994–4004

    2023

  8. [16]

    Diffusion facial forgery detection,

    H. Cheng, Y . Guo, T. Wang, L. Nie, and M. Kankanhalli, “Diffusion facial forgery detection,” inProceedings of the 32nd ACM International Conference on Multimedia (MM), 2024, p. 5939–5948

    2024

  9. [17]

    Genface: A large-scale fine-grained face forgery benchmark and cross appearance-edge learning,

    Y . Zhang, Z. Yu, T. Wang, X. Huang, L. Shen, Z. Gao, and J. Ren, “Genface: A large-scale fine-grained face forgery benchmark and cross appearance-edge learning,”IEEE Transactions on Information Forensics and Security, vol. 19, pp. 8559–8572, 2024

    2024

  10. [18]

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

    A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weissenborn, X. Zhai, T. Unterthiner, M. Dehghani, M. Minderer, G. Heigold, S. Gelly, J. Uszkoreit, and N. Houlsby, “An image is worth 16x16 words: Transformers for image recognition at scale,” inProceedings of the International Conf...

    2021

  11. [19]

    Fairadapter: Detecting ai- generated images with improved fairness,

    F. Ding, J. Zhang, X. He, and J. Xu, “Fairadapter: Detecting ai- generated images with improved fairness,” inICASSP 2025-2025 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2025, pp. 1–5

    2025

  12. [20]

    Forensics adapter: Adapting clip for generalizable face forgery detection,

    X. Cui, Y . Li, A. Luo, J. Zhou, and J. Dong, “Forensics adapter: Adapting clip for generalizable face forgery detection,” inProceedings of the Computer Vision and Pattern Recognition Conference, 2025, pp. 19 207–19 217

    2025

  13. [21]

    Frequency-aware deepfake detection: Improving generalizability through frequency space domain learning,

    C. Tan, Y . Zhao, S. Wei, G. Gu, P. Liu, and Y . Wei, “Frequency-aware deepfake detection: Improving generalizability through frequency space domain learning,” inProceedings of the AAAI Conference on Artificial Intelligence (AAAI), vol. 38, no. 5, 2024, pp. 5052–5060. JOURNAL ...

    2024

  14. [22]

    Im- plicit identity driven deepfake face swapping detection,

    B. Huang, Z. Wang, J. Yang, J. Ai, Q. Zou, Q. Wang, and D. Ye, “Im- plicit identity driven deepfake face swapping detection,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2023, pp. 4490–4499

    2023

  15. [23]

    Standing on the shoulders of giants: reprogramming visual-language model for general deepfake detection,

    K. Lin, Y . Lin, W. Li, T. Yao, and B. Li, “Standing on the shoulders of giants: reprogramming visual-language model for general deepfake detection,” inProceedings of the Thirty-Ninth AAAI Conference on Artificial Intelligence. AAAI Press, 2025

    2025

  16. [24]

    Eth-xgaze: A large scale dataset for gaze estimation under extreme head pose and gaze variation,

    X. Zhang, S. Park, T. Beeler, D. Bradley, S. Tang, and O. Hilliges, “Eth-xgaze: A large scale dataset for gaze estimation under extreme head pose and gaze variation,” inEuropean Conference on Computer Vision (ECCV), 2020

    2020

  17. [25]

    Multi- attentional deepfake detection,

    H. Zhao, W. Zhou, D. Chen, T. Wei, W. Zhang, and N. Yu, “Multi- attentional deepfake detection,” inProceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition (CVPR), June 2021, pp. 2185–2194

    2021

  18. [26]

    Lora: Low-rank adaptation of large language models

    E. J. Hu, Y . Shen, P. Wallis, Z. Allen-Zhu, Y . Li, S. Wang, L. Wang, W. Chenet al., “Lora: Low-rank adaptation of large language models.” ICLR, vol. 1, no. 2, p. 3, 2022

    2022

  19. [27]

    Fairface: Face attribute dataset for balanced race, gender, and age for bias measurement and mitigation,

    K. Karkkainen and J. Joo, “Fairface: Face attribute dataset for balanced race, gender, and age for bias measurement and mitigation,” inProceed- ings of the IEEE/CVF winter conference on applications of computer vision, 2021, pp. 1548–1558

    2021

  20. [28]

    Deep residual learning for image recognition,

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

    2016

  21. [29]

    Deepfake network architecture attribution,

    T. Yang, Z. Huang, J. Cao, L. Li, and X. Li, “Deepfake network architecture attribution,” inProceedings of the AAAI Conference on Artificial Intelligence (AAAI), 2022, pp. 4662–4670

    2022

  22. [30]

    Omnidfa: A unified framework for open set synthesis image detection and few-shot attribution,

    S. Wu, S. Li, J. Li, J. Liu, and Y . Wang, “Omnidfa: A unified framework for open set synthesis image detection and few-shot attribution,”arXiv preprint arXiv:2509.25682, 2025

    2025 arXiv

  23. [31]

    Learning counterfactually decoupled attention for open-world model attribution,

    Y . Zheng, B. Gong, F. Kong, Y . Duan, B. Yu, W. Zheng, L. Chen, J. Lu, and J. Zhou, “Learning counterfactually decoupled attention for open-world model attribution,” inICCV, 2025

    2025

  24. [32]

    De-fake: Detection and attri- bution of fake images generated by text-to-image generation models,

    Z. Sha, Z. Li, N. Yu, and Y . Zhang, “De-fake: Detection and attri- bution of fake images generated by text-to-image generation models,” inProceedings of the ACM SIGSAC Conference on Computer and Communications Security, 2023, p. 3418–3432

    2023

  25. [33]

    Improving fairness in deepfake detection,

    Y . Ju, S. Hu, S. Jia, G. H. Chen, and S. Lyu, “Improving fairness in deepfake detection,” inProceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2024, pp. 4655– 4665

    2024

  26. [34]

    Wilddeepfake: A challenging real-world dataset for deepfake detection,

    B. Zi, M. Chang, J. Chen, X. Ma, and Y .-G. Jiang, “Wilddeepfake: A challenging real-world dataset for deepfake detection,” inProceedings of the 28th ACM International Conference on Multimedia, 2020, pp. 2382–2390

    2020

  27. [35]

    Celeb-df: A large- scale challenging dataset for deepfake forensics,

    Y . Li, X. Yang, P. Sun, H. Qi, and S. Lyu, “Celeb-df: A large- scale challenging dataset for deepfake forensics,” in2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2020, pp. 3204–3213

    2020

  28. [36]

    The deepfake detection challenge dataset,

    B. Dolhansky, J. Bitton, B. Pflaum, J. Lu, R. Howes, M. Wang, and C. Canton-Ferrer, “The deepfake detection challenge dataset,”arXiv preprint arXiv:2006.07397, 2020

    2006 arXiv

  29. [37]

    DeeperForensics-1.0: A large-scale dataset for real-world face forgery detection,

    L. Jiang, R. Li, W. Wu, C. Qian, and C. C. Loy, “DeeperForensics-1.0: A large-scale dataset for real-world face forgery detection,” inProceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), June 2020, pp. 2886–2895

    2020

  30. [38]

    Bridging the gaps: Utilizing unlabeled face recognition datasets to boost semi-supervised facial expression recognition,

    J. Song, M. He, J. Feng, and B. Shen, “Bridging the gaps: Utilizing unlabeled face recognition datasets to boost semi-supervised facial expression recognition,”Neurocomputing, p. 133796, 2026

    2026

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.