Pith. sign in

REVIEW 3 major objections 5 minor 31 references

Unleashing the Potential of Vision-Language Models for Generalizable AI-Generated Image Detection

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

Pith's one-line read A frozen vision-language encoder becomes a state-of-the-art AI-image detector when its linear head starts from text embeddings of 'AI art' and 'a real photo'.

desk verdict A cheap and useful empirical win on AIGI detection whose mechanism story is confounded by learning rate and bias init; should go to review with a demand for a clean control. read the letter →

arxiv 2608.04935 v2 pith:7CKSU64K submitted 2026-08-05 cs.CV

classification cs.CV
keywords AI-generatedimagedetectionvision-languagemodelsPerceptionEncodersemanticprototypecalibrationlinearprobingforensicsemanticscross-generatorgeneralizationfrozenrepresentations
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 show that a frozen vision-language model, not a larger vision-only model, is the right base for detecting AI-generated images, and that the missing piece is how the classifier head is initialized. Under standard linear probing, the Perception Encoder (PE) underperforms DINOv3 by 4.1% on in-the-wild data even though its features separate real from generated images more cleanly. The proposed Semantic Prototype Calibration (SPC) starts the two columns of the binary head at normalized text embeddings of 'AI art' and 'a real photo', then calibrates them on a single supervised source. With this change, the same frozen encoder reaches 98.1% on GenImage, 94.0% on In-the-Wild, 95.2% on AIGI-Now, 98.7% on AIGI-Holmes, and 95.2% on Blur-and-JPEG, matching or beating the 6.7B-parameter DINOv3 baseline while using a 1.9B backbone. If true, this reframes generalizable AI-generated image detection as a head-initialization problem rather than a representation-scaling problem.

What carries the argument

The mechanism is Semantic Prototype Calibration (SPC). It treats the weight matrix $W=[p_1,p_2]$ of the binary linear head as two learnable category prototypes, initializes them with normalized text-encoder outputs for 'AI art' and 'a real photo' (Eq. 3), and then calibrates only the head on the frozen image encoder's features. The claim is that these text-derived prototypes sit near task-optimal directions in PE's embedding space, so calibration needs only small adjustments, preserving transferable forensic structure instead of overfitting the single training source.

What would settle it

Compare PE-SPC against PE-Linear using a Perception Encoder checkpoint whose pretraining data provably contained no AI-generated images; the paper's mechanism predicts a loss, not a gain, so a gain under SPC would falsify the claimed dependence on forensic pretraining exposure. A second check is to initialize the prototypes with text embeddings from a different text encoder and see whether the calibrated head still works, which would show whether PE-specific alignment is the load-bearing factor.

Watch

Extended reading notes

Core claim

The central discovery is that language-aligned representations already encode provenance semantics that linear probing wastes. On four in-the-wild collections, PE's frozen features have a pooled LDA overlap of 1.1% versus 12.3% for DINOv3, yet PE-Linear scores 89.9% versus DINOv3-Linear's 94.0%. SPC closes this gap by initializing the prototype columns $p_1,p_2$ as $p_1 \leftarrow N(f_T(\text{``AI art''}))$, $p_2 \leftarrow N(f_T(\text{``a real photo''}))$ and calibrating on the GenImage SD1.4 subset at a low learning rate. Ablations show equivalent prompts reproduce the gain, swapped or unrelated prompts destroy it, and models whose pretraining corpora predate widespread AI-generated imagery degrade under SPC. The paper concludes that the gain comes from class-aligned forensic semantics learned during pretraining, not from prompt wording or from text initialization per se.

Load-bearing premise

Everything rests on the assumption that the frozen Perception Encoder's text embeddings of 'AI art' and 'a real photo' already point near the directions in its image-feature space that separate generated from real images, which in turn requires that its pretraining corpus contained AI-generated images paired with captions or metadata identifying them as such.

Editorial extensions

If this is right

  • PE-SPC matches or exceeds DINOv3-Linear on every benchmark group while using a 1.9B-parameter backbone instead of 6.7B, so generalizable AI-image detection can run at roughly half the inference memory and about 2.7 times the throughput.
  • The method's gains concentrate on the hardest transfer settings, including ADM and Midjourney on GenImage, WildRF and SocialRF in the wild, and strong Gaussian blur, raising minimum accuracy and making detector behavior more stable.
  • SPC transfers to other vision-language models only when pretraining exposed them to AI-generated imagery with provenance-revealing captions; MetaCLIP2 improves, while OpenCLIP and SigLIP2 degrade.
  • Because the text encoder is used only at initialization, PE-SPC adds no inference-time cost over a linear probe on the same frozen encoder.

Reading between the lines

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

  • If SPC's mechanism is general, the same seed-from-text recipe may apply to other binary provenance questions, such as deepfake video, synthetic audio, or document provenance, wherever a vision-language model has seen the two classes described in text.
  • A testable extension is to measure SPC's gain as a function of pretraining corpus date or AI-image exposure; the paper's OpenCLIP and SigLIP2 comparison predicts a monotone relationship that a controlled corpus-ablation experiment could verify.
  • The two fixed prompts are probably not optimal; learning prompt embeddings or using per-generator prompt pools would likely extract additional forensic signal from PE, a direction the paper does not explore.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The manuscript proposes Semantic Prototype Calibration (SPC) for AI-generated-image (AIGI) detection. SPC initializes the two columns of a binary linear head with normalized text embeddings of "AI art" and "a real photo" from the frozen Perception Encoder (PE) text encoder, and then calibrates these prototypes on the GenImage SD1.4 training set with supervised data. The authors report consistent gains over PE-Linear across five benchmark groups---GenImage, In-the-Wild, AIGI-Now, AIGI-Holmes, and Blur-and-JPEG---and show that PE-SPC matches or exceeds DINOv3-Linear, a much larger vision-only model, while using the 1.9B-parameter PE instead of the 6.7B-parameter DINOv3. The paper also analyzes the source of the gains, arguing that the text-derived prototypes provide forensic semantic starting points that lie near task-optimal directions, and presents prompt ablations and a cross-model comparison to support this mechanism.

Significance. If the central claim holds, this is a meaningful result: it shows that a simple text-initialized linear probe on a frozen vision-language model can surpass a substantially larger vision-only baseline on a practical forensic task, and it offers a concrete, testable explanation in terms of pretraining exposure to AI-generated imagery. The manuscript has real strengths: it evaluates on multiple diverse benchmarks, includes per-dataset feature-space analyses, provides prompt and model ablations, and reports efficiency measurements. The prompt ablation design (Table 6) is thoughtful and helps separate wording from semantics. The main weaknesses are experimental rather than conceptual: the key comparison between PE-Linear and PE-SPC changes optimization settings at the same time as the initialization, the learning rate is selected using the same benchmarks that are then reported as headline results, and no variance information is given. These issues are load-bearing because the manuscript's central claim is that text-derived forensic semantics, not a generic optimization change, drive the improvement.

major comments (3)
  1. [Implementation Details and Eq. (3)] The comparison between PE-Linear and PE-SPC differs in three variables at once: the weight initialization (random vs. text embeddings of Eq. (3)), the learning rate (1e-3 vs. 2e-5), and the bias initialization (default vs. 1/0). The paper attributes the gain entirely to the semantic starting point, but no experiment controls for the optimization changes. A randomly initialized head trained under the exact SPC optimizer settings (learning rate 2e-5, bias 1/0) is needed, as is a head initialized with random normalized vectors under the same settings. Table 6 compares prompt contents while keeping the SPC-specific optimizer fixed, but it does not include a random-initialization control; it therefore cannot by itself separate semantic content from optimization dynamics. This is not a peripheral issue: Appendix C shows that the SPC-specific learning rate is important for the In-the-Wild result, so without the random-init control the headline gains could be an artifact of learning rate and bias initialization rather than of forensic semantics.
  2. [Appendix C and Figure 5] The learning rate 2e-5 was selected by evaluating all five benchmark groups that are then used for the headline comparisons. This makes the reported numbers optimistic in a way that standard peer review does not account for: the same data were used both to choose a hyperparameter and to report the final evaluation. A separate validation split, or at least one deferred benchmark group, should be used to select the learning rate, and the selection rule should be described. This point is load-bearing because several of the reported improvements over DINOv3-Linear are small (e.g., +1.7% on GenImage, +1.5% on AIGI-Holmes), and Appendix C indicates that a different learning rate would change the In-the-Wild accuracy substantially.
  3. [Tables 1-5] No error bars, standard deviations, or significance tests are reported anywhere in the main results. Most entries are point estimates from a single run, and some differences that the narrative relies on are modest. Given that only a few thousand parameters are trained for two epochs, seed sensitivity is a realistic concern. The authors should report repeated-seed means with standard deviations, or at least provide bootstrapped confidence intervals on the accuracy differences, so that the reader can judge whether values such as 98.1% vs. 96.4% or 95.2% vs. 94.0% are distinguishable. This is particularly relevant for the claim that PE-SPC 'matches or exceeds' DINOv3-Linear, which partly rests on differences of less than two percentage points.
minor comments (5)
  1. [Method / Figure 1(b) and Appendix E] The LDA analysis uses a random sample of features from each dataset, but the number of sampled features and the random seed are not given. Adding these details would make the motivation reproducible.
  2. [Appendix C, Figure 5] The vertical axis of Figure 5 starts at 50%, which visually compresses the differences between learning rates. Starting the axis at 0, or explicitly noting the truncation in the caption, would make the sensitivity plot less misleading.
  3. [Main text table references] The table references are inconsistent: the text says 'In Tables 3' and 'in table 4' instead of 'in Table 3' and 'Table 4'. Please normalize all references to the same format.
  4. [Results / Table 6 caption] The caption of Table 6 says 'Both improve all five groups' but the 'Similar Prompt' row shows a small decrease on In-the-Wild relative to 'Our Prompt' (90.3% vs. 94.0%); the sentence should be rephrased to say that both prompts improve over the 'No SPC' baseline, not that they improve every group relative to each other.
  5. [Appendix D, Table 8] The pretraining-data column reports publication years rather than actual data-collection cutoffs, and the table footnote says the years 'indicate publication years rather than exact data-collection cutoffs.' This caveat is helpful, but the main text in Appendix D sometimes speaks as if LAION and WebLI were collected before the growth of synthetic content; please align the wording with the table's caveat.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: SPC's gains are measured on external benchmarks, and the semantic-prototype explanation rests on post-hoc interpretation rather than a derivation that equals its input.

full rationale

The derivation chain is not circular. SPC initializes the two linear-head columns with normalized frozen text embeddings of 'AI art' and 'a real photo' (Eq. 3), then calibrates them on GenImage SD1.4; the reported accuracies come from held-out benchmark groups (GenImage other generators, In-the-Wild, AIGI-Holmes, AIGI-Now, Blur-and-JPEG). None of these quantities is defined in terms of the output it is said to predict, and no fitted parameter is renamed as a prediction. The claim that text-derived prototypes 'lie near task-optimal directions' is an empirical explanation inferred from the same benchmark gains, and Appendix C shows the learning rate was selected by looking at all five benchmark groups; the comparison PE-SPC versus PE-Linear also changes learning rate and bias initialization simultaneously. These are evaluation-protocol and attribution concerns, not circularity: the numbers are not forced by construction, and the prompt ablations in Table 6 provide external (if incomplete) evidence that class-aligned semantic content matters. No load-bearing self-citation is used; the cited prior results (PE, DINOv3, Simplicity Prevails) are external model papers and benchmarks. Score 1 reflects a minor post-hoc-explanation concern rather than any actual circular reduction.

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

The central claim rests on three unproved premises: the alignment of text prototypes with PE's image space, the forensic content of PE's pretraining corpus, and the transferability of frozen VFM features. The paper gives empirical support for all three, but none is independently verified outside its own evaluation protocol.

free parameters (3)
  • SPC learning rate = 2e-5
    Chosen by sweeping over the five benchmark groups and selecting the best balance (Appendix C, Figure 5); this is evaluation-set tuning.
  • Text prompt pair for prototypes = 'AI art' and 'a real photo'
    Hand-picked forensic descriptions; the paper's ablation (Table 6) shows wording matters only through semantic alignment, but the specific choices are the method's starting points.
  • Bias initialization = generated=1, real=0
    Chosen to stabilize early calibration (Implementation details), not derived from data.
assumptions (3)
  • ad hoc to paper PE's text embeddings of 'AI art' and 'a real photo' lie near task-optimal directions in its frozen image feature space.
    Eq. (3) and the section 'Semantic Prototype Calibration'; if false, SPC degrades, as shown by the prompt-swap ablations in Table 6.
  • domain assumption PE was pretrained on a web corpus containing large volumes of AI-generated images paired with captions or metadata identifying them as AI-generated.
    Inferred from model papers and Appendix D; the paper does not verify corpus composition directly. Figure 4's contrast with OpenCLIP and SigLIP2 supports it indirectly.
  • domain assumption Frozen VFM features contain transferable forensic provenance information sufficient for AIGI detection.
    Assumed from prior work (Simplicity Prevails, SSAFE), cited rather than re-derived; the entire linear-probing paradigm depends on it.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Unleashing the Potential of Vision-Language Models for Generalizable AI-Generated Image Detection." pith.science (2026). https://pith.science/paper/7CKSU64K

@misc{pith2026260804935,
  author       = {Pith},
  title        = {Pith review of: Unleashing the Potential of Vision-Language Models for Generalizable AI-Generated Image Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7CKSU64K}},
  note         = {Machine review of arXiv:2608.04935}
}
read the original abstract

Recent work has shown that a simple linear probe on frozen representations from modern vision foundation models (VFMs) can achieve state-of-the-art AIGI detection performance, substantially outperforming specialized detectors in challenging in-the-wild scenarios. This finding has established DINOv3 as the dominant foundation-model baseline for subsequent improvements. However, we find that the vision-language model Perception Encoder (PE) holds greater potential for AIGI detection, because its language-aligned representation preserves high-level provenance semantics. Specifically, PE exhibits stronger local provenance organization than DINOv3 in its frozen feature space. However, semantic-agnostic linear probing fails to exploit this structure, as PE-Linear still underperforms DINOv3-Linear by 4.1% on In-the-Wild. Based on this observation, we propose Semantic Prototype Calibration (SPC), which constructs category prototypes from forensic semantic information and calibrates them with supervised data. We apply SPC to PE and refer to the resulting detector as PE-SPC. Our analysis shows that this simple design achieves stronger generalization. Across cross-generator, post-processing, and in-the-wild benchmarks, PE-SPC surpasses the previous DINOv3 baseline and achieves new state-of-the-art results.

Figures

Figures reproduced from arXiv: 2608.04935 by the authors.

Figure 1
Figure 1. Our motivation is that standard linear probing fails to fully exploit the forensic structure already present in PE’s features. [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Category-prototype view of linear probing and Se [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Average and minimum accuracy of PE-Linear, PE-SPC, and DINOv3-Linear across five benchmark groups. Bar [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Effect of SPC on vision-language models with [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Learning-rate sensitivity of PE-SPC across the five [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]
Figure 6
Figure 6. Figure 6: Per-dataset feature-space analysis of DINOv3 and PE, complementing the pooled analysis in Figure 1(b). Each [PITH_FULL_IMAGE:figures/full_fig_p011_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

31 extracted references · 22 canonical work pages

  1. [1]

    Abdullah, A.; Ebert, N.; and Wasenm \"u ller, O. 2026. TAP into the Patch Tokens: Leveraging Vision Foundation Model Features for AI -Generated Image Detection. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 379--389

  2. [2]

    H.; Madotto, A.; Wei, C.; Ma, T.; Zhi, J.; Rajasegaran, J.; Bangalath, H.; Wang, J.; Monteiro, M.; Xu, H.; Dong, S.; Ravi, N.; Li, S.-W.; Doll \'a r, P.; and Feichtenhofer, C

    Bolya, D.; Huang, P.-Y.; Sun, P.; Cho, J. H.; Madotto, A.; Wei, C.; Ma, T.; Zhi, J.; Rajasegaran, J.; Bangalath, H.; Wang, J.; Monteiro, M.; Xu, H.; Dong, S.; Ravi, N.; Li, S.-W.; Doll \'a r, P.; and Feichtenhofer, C. 2025. Perception Encoder: The Best Visual Embeddings Are Not at the Output of the Network. In Advances in Neural Information Processing Sys...

  3. [3]

    Cavia, B.; Horwitz, E.; Reiss, T.; and Hoshen, Y. 2024. Real-Time Deepfake Detection in the Real-World. arXiv preprint arXiv:2406.09398

  4. [4]

    Chen, R.; Gao, J.; Lin, K.; Zhang, K.; Zhao, Y.; Guan, I.; Yao, T.; and Ding, S. 2025 a . Task-Model Alignment: A Simple Path to Generalizable AI -Generated Image Detection. arXiv preprint arXiv:2512.06746

  5. [5]

    Chen, R.; Xi, J.; Yan, Z.; Zhang, K.-Y.; Wu, S.; Xie, J.; Chen, X.; Xu, L.; Guan, I.; Yao, T.; and Ding, S. 2025 b . Dual Data Alignment Makes AI -Generated Image Detector Easier Generalizable. In Advances in Neural Information Processing Systems, volume 38

  6. [6]

    Cherti, M.; Beaumont, R.; Wightman, R.; Wortsman, M.; Ilharco, G.; Gordon, C.; Schuhmann, C.; Schmidt, L.; and Jitsev, J. 2023. Reproducible Scaling Laws for Contrastive Language-Image Learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2818--2829

  7. [7]

    E.; Zettlemoyer, L.; Chen, X.; Liu, Z.; Xie, S.; Yih, S.; Li, S.-W.; and Xu, H

    Chuang, Y.-S.; Li, Y.; Wang, D.; Yeh, C.-F.; Lyu, K.; Raghavendra, R.; Glass, J.; Huang, L.; Weston, J. E.; Zettlemoyer, L.; Chen, X.; Liu, Z.; Xie, S.; Yih, S.; Li, S.-W.; and Xu, H. 2025. Meta CLIP 2: A Worldwide Scaling Recipe. In Advances in Neural Information Processing Systems, volume 38. Spotlight presentation

  8. [8]

    Fisher, R. A. 1936. The Use of Multiple Measurements in Taxonomic Problems. Annals of Eugenics, 7(2): 179--188

Show all 31 references
  1. [9]

    Huang, Z.; Li, J.; Wen, H.; Li, T.; Yang, X.; Qi, L.; Peng, B.; Huang, X.; Yang, M.-H.; and Cheng, G. 2025. Rethinking Cross-Generator Image Forgery Detection through DINOv3 . arXiv preprint arXiv:2511.22471

  2. [10]

    F.; and Bradley, E

    Inman, H. F.; and Bradley, E. L., Jr. 1989. The Overlapping Coefficient as a Measure of Agreement between Probability Distributions and Point Estimation of the Overlap of Two Normal Densities. Communications in Statistics---Theory and Methods, 18(10): 3851--3874

  3. [11]

    Lee, S.; Kim, B.; Nam, J.; Lee, K.; and Shin, J. 2026. SSAFE : Simple and Strong AI -Generated Image Detection via Frozen Vision Encoders. arXiv preprint arXiv:2606.08634

  4. [12]

    Li, X.; Zhang, J.; Li, Y.; Cao, Y.; and Wang, W. 2026. Spectral Tail Auxiliary Learning for AI -Generated Image Detection. arXiv preprint arXiv:2605.22751

  5. [13]

    Li, Z.; Yan, J.; He, Z.; Zeng, K.; Jiang, W.; Xiong, L.; and Fu, Z. 2025. Is Artificial Intelligence Generated Image Detection a Solved Problem? In Advances in Neural Information Processing Systems, volume 38. Datasets and Benchmarks Track

  6. [14]

    Liu, W.; Wen, Y.; Yu, Z.; Li, M.; Raj, B.; and Song, L. 2017. SphereFace : Deep Hypersphere Embedding for Face Recognition. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition

  7. [15]

    Ojha, U.; Li, Y.; and Lee, Y. J. 2023. Towards Universal Fake Image Detectors That Generalize Across Generative Models. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 24480--24489

  8. [16]

    W.; Hallacy, C.; Ramesh, A.; Goh, G.; Agarwal, S.; Sastry, G.; Askell, A.; Mishkin, P.; Clark, J.; Krueger, G.; and Sutskever, I

    Radford, A.; Kim, J. W.; Hallacy, C.; Ramesh, A.; Goh, G.; Agarwal, S.; Sastry, G.; Askell, A.; Mishkin, P.; Clark, J.; Krueger, G.; and Sutskever, I. 2021. Learning Transferable Visual Models From Natural Language Supervision. In Meila, M.; and Zhang, T., eds., Proceedings of...

  9. [17]

    Schuhmann, C.; Beaumont, R.; Vencu, R.; Gordon, C.; Wightman, R.; Cherti, M.; Coombes, T.; Katta, A.; Mullis, C.; Wortsman, M.; Schramowski, P.; Kundurthy, S.; Crowson, K.; Schmidt, L.; Kaczmarczyk, R.; and Jitsev, J. 2022. LAION-5B : An Open Large-Scale Dataset for Training N...

  10. [18]

    Shi, K.; Lu, J.; Ye, S.; Zhang, G.; and Fang, Z. 2025. MiraGe : Multimodal Discriminative Representation Learning for Generalizable AI -Generated Image Detection. In Proceedings of the 33rd ACM International Conference on Multimedia, 353--361

  11. [19]

    Sim \'e oni, O.; Vo, H. V.; Seitzer, M.; Baldassarre, F.; Oquab, M.; Jose, C.; Khalidov, V.; Szafraniec, M.; Yi, S.; Ramamonjisoa, M.; Massa, F.; Haziza, D.; Wehrstedt, L.; Wang, J.; Darcet, T.; Moutakanni, T.; Sentana, L.; Roberts, C.; Vedaldi, A.; Tolan, J.; Brandt, J.; Coup...

  12. [20]

    Tan, C.; Tao, R.; Liu, H.; Gu, G.; Wu, B.; Zhao, Y.; and Wei, Y. 2025. C2P-CLIP : Injecting Category Common Prompt in CLIP to Enhance Generalization in Deepfake Detection. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 39, 7184--7192

  13. [21]

    F.; Alabdulmohsin, I.; Parthasarathy, N.; Evans, T.; Beyer, L.; Xia, Y.; Mustafa, B.; H \'e naff, O.; Harmsen, J.; Steiner, A.; and Zhai, X

    Tschannen, M.; Gritsenko, A.; Wang, X.; Naeem, M. F.; Alabdulmohsin, I.; Parthasarathy, N.; Evans, T.; Beyer, L.; Xia, Y.; Mustafa, B.; H \'e naff, O.; Harmsen, J.; Steiner, A.; and Zhai, X. 2025. SigLIP 2: Multilingual Vision-Language Encoders with Improved Semantic Understan...

  14. [22]

    Wang, F.; Xiang, X.; Cheng, J.; and Yuille, A. L. 2017. NormFace : L_2 Hypersphere Embedding for Face Verification. In Proceedings of the 25th ACM International Conference on Multimedia, 1041--1049

  15. [23]

    Wang, H.; Wang, Y.; Zhou, Z.; Ji, X.; Gong, D.; Zhou, J.; Li, Z.; and Liu, W. 2018. CosFace : Large Margin Cosine Loss for Deep Face Recognition. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition

  16. [24]

    Xu, H.; Chen, Z.; Zhang, H.; Xue, L.; and Zhang, H. 2025. GCS-Net : A Universal AI -Generated Visual Content Detection Method Based on CLIP . Knowledge-Based Systems, 323: 113806

  17. [25]

    Yan, S.; Li, O.; Cai, J.; Hao, Y.; Jiang, X.; Hu, Y.; and Xie, W. 2025. A Sanity Check for AI -Generated Image Detection. In International Conference on Learning Representations

  18. [26]

    Yan, Z.; Wang, J.; Jin, P.; Zhang, K.-Y.; Liu, C.; Chen, S.; Yao, T.; Ding, S.; Wu, B.; and Yuan, L. 2024. Orthogonal Subspace Decomposition for Generalizable AI -Generated Image Detection. arXiv preprint arXiv:2411.15633

  19. [27]

    Zhou, Y.; He, X.; Lin, K.; Fan, B.; Ding, F.; and Li, B. 2025 a . Breaking Latent Prior Bias in Detectors for Generalizable AIGC Image Detection. In Advances in Neural Information Processing Systems, volume 38

  20. [28]

    Zhou, Y.; He, X.; Lin, K.; Fan, B.; Ding, F.; and Li, B. 2026. Simplicity Prevails: The Emergence of Generalizable AIGI Detection in Visual Foundation Models. arXiv preprint arXiv:2602.01738

  21. [29]

    Zhou, Y.; He, X.; Lin, K.; Fan, B.; Ding, F.; Zeng, J.; and Li, B. 2025 b . Brought a Gun to a Knife Fight: Modern VFM Baselines Outgun Specialized Detectors on In-the-Wild AI Image Detection. arXiv preprint arXiv:2509.12995

  22. [30]

    Zhou, Z.; Luo, Y.; Wu, Y.; Sun, K.; Ji, J.; Yan, K.; Ding, S.; Sun, X.; Wu, Y.; and Ji, R. 2025 c . AIGI-Holmes : Towards Explainable and Generalizable AI -Generated Image Detection via Multimodal Large Language Models. In Proceedings of the IEEE/CVF International Conference o...

  23. [31]

    Zhu, M.; Chen, H.; Yan, Q.; Huang, X.; Lin, G.; Li, W.; Tu, Z.; Hu, H.; Hu, J.; and Wang, Y. 2023. GenImage : A Million-Scale Benchmark for Detecting AI -Generated Image. In Advances in Neural Information Processing Systems, volume 36, 77771--77782

Pith tools

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