Pith. sign in

REVIEW 4 major objections 5 minor 30 references

PatchHead: Learning Spatial Patch Evidence for Generalizable AI-Generated Image 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 Preserving the spatial layout of DINO patch tokens improves cross-dataset detection of AI-generated images.

desk verdict Useful hypothesis and honest ablations, but the headline gain is mostly protocol, not the spatial head; still worth a serious referee. read the letter →

arxiv 2608.09223 v1 pith:34LA5DCR submitted 2026-08-10 cs.CV

Shengbo Qi , Hongyi Fang , Benjia Zhou , Rui Mao This is my paper · ORCID
classification cs.CV
keywords AI-generatedimagedetectioncross-datasetgeneralizationDINOpatchtokensspatialaggregationCLStokenLoRAfine-tuningdeepfakeout-of-distributionrobustness
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

AI-generated image detectors that rely on a single global CLS token discard the spatial layout of image patches. This paper argues that generation traces are local and spatially distributed, so the detector interface should preserve patch positions and combine neighboring evidence before making a decision. PatchHead does this with a small convolutional head placed after a frozen DINOv3 backbone, and the paper reports that it raises average cross-dataset balanced accuracy from 91.6% to 94.6% and worst-case accuracy from 82.4% to 89.4% across nine benchmarks. The point is that how foundation-model features are aggregated matters as much as which features are used.

What carries the argument

The central object is PatchHead, a lightweight spatial aggregation head. It takes the sequence of DINO patch tokens, reshapes it to a two-dimensional feature map, applies a $3\times3$ depthwise convolution to exchange information among neighboring patches, a GELU activation, a $1\times1$ pointwise convolution to project channels, and global average pooling before a linear classifier; the CLS token is used only in an auxiliary contrastive loss during training and discarded at inference. This mechanism is what tests the paper's hypothesis because it isolates spatially structured aggregation against global CLS compression while keeping the backbone and training protocol fixed.

What would settle it

Retrain PatchHead on a source set built from generators unrelated to the DDA pipeline and evaluate on DDA-COCO and EvalGEN; if the average or worst-case gains over the CLS baseline fall below the claimed +3.0 and +6.9 points, part of the improvement is explained by training data overlap.

Watch

Extended reading notes

Core claim

The paper claims that the standard practice of classifying AI-generated images from the CLS token underuses the spatial representations of DINO-family backbones, and that restoring patch tokens to their native two-dimensional layout before aggregation yields a more transferable authenticity signal. Trained on the DDA-aligned COCO-SD-2 dataset with the backbone frozen and LoRA adapters learned, PatchHead reshapes the patch token sequence into a spatial feature map $F_0 \in \mathbb{R}^{D \times H_p \times W_p}$, applies a $3 \times 3$ depthwise convolution followed by a $1 \times 1$ projection, pools globally, and classifies from that pooled patch representation. The reported result is first or second place on all nine cross-dataset benchmarks, with an average balanced accuracy of 94.6% and a worst-case accuracy of 89.4%, improving on the strongest prior method by 3.0 and 6.9 points respectively while adding only 8.6% trainable parameters and 0.08% FLOPs. Ablations show that simple patch averaging does not reproduce the gain and that shuffling patch positions degrades it, which the paper interprets as evidence that learnable spatial aggregation over the native layout is the active ingredient.

Load-bearing premise

The reported gains assume all nine target benchmarks are genuinely unseen at training time, even though DDA-COCO and EvalGEN come from the same research line that produced the training data.

Editorial extensions

If this is right

  • A single trained PatchHead model, without target-specific fine-tuning or test-time adaptation, is competitive or best on datasets spanning GAN, diffusion, flow-matching, and in-the-wild sources.
  • Learnable spatial aggregation, not merely access to patch tokens, is required: global averaging of patches underperforms the CLS baseline.
  • Preserving the native 2D layout matters: a fixed random permutation of patch positions lowers average accuracy from 94.6% to 94.3%.
  • The spatial response analysis implies that detectors can be steered toward distributed authenticity evidence rather than content-salient regions, measurable in coverage and top-10% activation statistics.
  • Heavier spatial designs such as multi-scale, multi-level fusion, and HRM-inspired refinement do not automatically improve cross-domain accuracy; the simple depthwise-conv head is stronger.

Reading between the lines

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

  • Beyond the paper, if the representation-interface account is right, applying the same spatial-aggregation idea to other vision foundation models and to other forensic tasks should reproduce the transfer gain; this is directly testable.
  • Beyond the paper, the claim that spatial evidence is authenticity-sensitive suggests that combining PatchHead with reconstruction-based or frequency-domain cues could push worst-case accuracy further, since blur and heavy post-processing remain hard for both interfaces.
  • Beyond the paper, a stricter evaluation protocol that excludes DDA-COCO and EvalGEN from the target list would clarify how much of the improvement is due to spatial aggregation rather than to benchmark overlap with the DDA training pipeline.
  • Beyond the paper, PatchHead's stability under resizing hints that the spatial head is more robust to geometric perturbations; testing stronger JPEG compression or combined degradations would show the limits.
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 proposes PatchHead, a lightweight classification head for AI-generated image detection that replaces the conventional CLS-token interface of a frozen DINOv3 backbone with a spatial aggregation of patch tokens. The head reshapes patch tokens to their native 2D layout, applies a depthwise 3×3 convolution followed by a 1×1 projection and global average pooling, and trains end-to-end with only LoRA adapters plus an auxiliary contrastive loss on the CLS token, which is discarded at inference. The authors report cross-dataset balanced accuracy on nine benchmarks, claiming an average improvement of +3.0 points and a worst-case improvement of +6.9 points over the strongest prior method, DDA, while adding only 8.6% trainable parameters and 0.08% FLOPs. Controlled ablations compare PatchHead to CLS-based and GAP-based heads under the same training protocol, and additional analyses examine spatial response statistics, feature-space alignment, robustness to perturbations, and sensitivity to LoRA and contrastive-loss settings.

Significance. If the central attribution held, the paper would provide a simple and inexpensive interface change for foundation-model-based detectors, with careful single-model cross-dataset evaluation and a clear representation-level hypothesis. The controlled ablations in Table 4 are a genuine strength: they isolate the head choice under a shared backbone, data, augmentations, and objective, and the spatial-response and feature-space analyses offer falsifiable qualitative evidence that spatial aggregation changes where the detector attends. The architecture is simple, the overhead is small, and the nine-benchmark protocol without target adaptation is a useful contribution. However, the headline quantitative claims are not yet supported as stated, because the main comparison is not controlled for the backbone/training protocol and the worst-case margin is computed over different benchmarks.

major comments (4)
  1. [Table 1 vs. Table 4 (Experiments, Comparison with State-of-the-Art Methods)] The headline claim that PatchHead improves the strongest prior method by +3.0 average and +6.9 worst-case points is not a controlled comparison for the paper's causal claim. The DDA row in Table 1 is the prior system from Chen et al. 2025, which uses a different backbone and training protocol, whereas Table 4's same-backbone CLSHead(Linear) baseline differs from PatchHead by only +0.7 average and +0.5 minimum points on the same nine benchmarks. The large deltas in Table 1 are therefore largely attributable to the DINOv3+LoRA+contrastive protocol rather than to spatial aggregation, so the central statement that spatial patch aggregation transfers more reliably is not supported by the headline numbers. Please report the head-only comparison as the primary quantitative evidence, or provide a same-protocol DDA-equivalent baseline.
  2. [Table 1 and Table 3 (Worst-case accuracy)] The reported worst-case improvement of +6.9 points compares minima over different benchmarks: DDA's minimum (82.4) occurs on Chameleon, while PatchHead's minimum (89.4) occurs on SynthWildX, where PatchHead is actually below DDA (89.4 vs. 90.9). This per-benchmark minimum comparison inflates the robustness gain. Please report paired per-benchmark differences, the average of per-benchmark deltas, and the minimum per-benchmark delta, and avoid claiming a 'worst-case gain' unless the minima are computed on the same benchmark.
  3. [Figure 6 and Equation (8)] The contrastive-loss weight λ=0.5 is selected via ablation in Figure 6 on the same nine evaluation benchmarks that are then used for the final reported numbers, and no validation split or multiple-seed error bars are provided. Because each ablation appears to be a single training run, selection on the evaluation set combined with the absence of variance estimates makes the reported margins, especially the 0.7-point gap in Table 4, statistically ungrounded. Please describe the selection procedure with a held-out split, report error bars over at least three seeds, and show the sensitivity of the Table 4 comparison to λ.
  4. [Datasets and Appendix B] The claim that the nine benchmarks are unseen targets is not fully established for DDA-COCO and EvalGEN. Both come from Chen et al. 2025, the same source as the DDA-aligned COCO-SD-2 training set, and the paper does not demonstrate that their generator sources or alignment artifacts are disjoint from the training distribution. If these two benchmarks share distributional components with the training data, the average and worst-case gains in Table 1 are inflated. Please provide a disjointness analysis, for example a comparison of generator lists, alignment transforms, or a leakage probe.
minor comments (5)
  1. [Table 3] The asterisks on the SAFE and AIDE rows (54.2 and 54.7) are unexplained; specify what values were corrected and why, matching the Table 1 footnote convention.
  2. [Table 4 and Abstract] The FLOPs column in Table 4 reports values such as '~0.45M' and '~242M', while the abstract states '0.08% additional FLOPs'; clarify the units, the reference point for the percentage, and how the head-only FLOPs are measured.
  3. [Figure 6] The plot is difficult to read in grayscale; use distinct markers and, if possible, show per-seed points or error bars to support the claimed insensitivity to λ.
  4. [Appendix B] The supplement states that 'detailed statistics are provided in the supplementary material' for the nine benchmarks, but the full text only contains Table 7; ensure every referenced supplementary figure and table is present and numbered consistently.
  5. [References] Several 2026 references (e.g., DINOv3, SimLBR, MIRROR) are used; please mark which are preprints and ensure their publication status is correctly described.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the spatial head is evaluated as an empirical architectural choice, not derived from its own assumptions.

full rationale

The paper's central claim is that replacing the CLS classifier with a spatial patch-aggregation head improves cross-dataset AI-generated-image detection. This is not a derivation but an empirical comparison, and the load-bearing controlled evidence (Table 4) is self-contained: all variants share the same DINOv3 backbone, LoRA configuration, training data, augmentations, and contrastive objective, so the +0.7/+0.5 head-only margins isolate the aggregation interface rather than reducing to a fitted parameter. The headline +3.0/+6.9 numbers versus DDA are not circular; they are an external comparison, though DDA is not a same-protocol re-implementation, which is a validity concern, not a circularity one. The only mild protocol issue is that lambda = 0.5 was selected by ablation on four of the nine evaluation benchmarks and then reported on those same benchmarks; this can inflate reported numbers, but it does not make the reported accuracy logically equivalent to lambda, and the head comparison is made at a fixed lambda. No equation in the paper reduces PatchHead's output to its training target, no load-bearing result is justified solely by a self-citation, and no uniqueness theorem or ansatz is imported from the authors' prior work. By the definitions of circularity used here, the derivation chain is self-contained; the weaknesses are benchmark-contamination and comparison-control issues, not circularity.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The paper's contribution rests on a small number of empirical assumptions rather than a mathematical derivation. The main free parameters are hyperparameters selected by ablation on the same benchmarks. There are no invented physical entities.

free parameters (3)
  • contrastive loss weight lambda = 0.5
    Selected by ablation in Fig. 6 on four of the evaluation benchmarks, then used for all reported results; not selected on a fully held-out validation set.
  • LoRA rank = 8
    Rank 8 with alpha 1 chosen as default; Table 12 shows nearby settings change individual results by about one point, so the choice affects the headline numbers modestly.
  • LoRA alpha = 1
    Alpha=4 gives 94.7 vs 94.6 average, but the default was kept for balance; this is a hand-selected hyperparameter.
assumptions (4)
  • domain assumption DINOv3 patch tokens carry spatially distributed authenticity-relevant traces that survive to the final layer.
    This is the core hypothesis motivating PatchHead; the paper argues for it via ablations and spatial response maps, but no causal proof is given.
  • domain assumption The DDA-aligned COCO-SD-2 training distribution is a suitable source domain for cross-generator generalization.
    All models are trained once on this dataset; if it is unrepresentative, the cross-dataset results would change.
  • domain assumption The nine benchmarks are valid disjoint target domains, including DDA-COCO and EvalGEN.
    Training and two of the nine test sets come from the same paper (Chen et al. 2025); no explicit generator-disjointness analysis is provided.
  • domain assumption Attention Rollout and CAM maps are comparable after min-max normalization.
    The spatial response statistics in Fig. 4 and Appendix D compare attribution maps produced by different mechanisms; this comparability is assumed, not established.

how reviews work

0 comments
Cite this review

Pith. "Pith review of PatchHead: Learning Spatial Patch Evidence for Generalizable AI-Generated Image Detection." pith.science (2026). https://pith.science/paper/34LA5DCR

@misc{pith2026260809223,
  author       = {Pith},
  title        = {Pith review of: PatchHead: Learning Spatial Patch Evidence for Generalizable AI-Generated Image Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/34LA5DCR}},
  note         = {Machine review of arXiv:2608.09223}
}
read the original abstract

AI-generated image detectors generalize poorly when their training and test images originate from different generators or datasets. Despite the rich spatial representations produced by vision foundation models like DINO, existing detectors typically classify images using only the globally aggregated CLS token. We hypothesize that globally aggregating DINO features into a single CLS token obscures spatially distributed generation traces. To test this hypothesis, we introduce PatchHead, a lightweight spatial aggregation head that preserves the two-dimensional organization of DINO patch tokens and integrates evidence across neighboring regions. During training, we freeze the pretrained DINO backbone and optimize only the inserted LoRA adapters, PatchHead, and auxiliary projection head. Across nine cross-dataset benchmarks spanning manually curated and in-the-wild settings, PatchHead ranks first on seven datasets and second on the remaining two. It improves the strongest prior method from 91.6% to 94.6% in average balanced accuracy (+3.0 points) and raises the worst-case accuracy from 82.4% to 89.4% (+6.9 points), while introducing only 8.6% more trainable parameters and 0.08% additional FLOPs. Further qualitative analysis suggests that PatchHead (i) reduces class-conditional domain discrepancy, and (ii) redirects the representation from content-dominated saliency toward spatially distributed authenticity evidence. Together, these observations provide a representation-level account of why spatial patch aggregation transfers more reliably across generators and datasets than a single CLS-based global representation. Our code and models will be made available upon acceptance.

Figures

Figures reproduced from arXiv: 2608.09223 by the authors.

Figure 1
Figure 1. Radar visualization of balanced accuracy on cross [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Conceptual comparison of representation strategies [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Overall framework of PatchHead. DINOv3 ViT-L/16 is adopted as the visual backbone and optimized with LoRA [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Qualitative and quantitative comparisons of spatial responses. Left: representative response maps of the CLS-based [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: t-SNE visualization of feature distributions learned by the CLS-based baseline and PatchHead on the training set [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: Effect of different contrastive learning weights for [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: Additional qualitative visualization of spatial responses. [PITH_FULL_IMAGE:figures/full_fig_p013_7.png]
Figure 8
Figure 8. Figure 8: Additional quantitative spatial response analysis on Chameleon and WildRF datasets. [PITH_FULL_IMAGE:figures/full_fig_p013_8.png]

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

30 extracted references · 22 canonical work pages

  1. [1]

    Abdullah, A.; Ebert, N.; and Wasenm\"uller, 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 (CVPR) Workshops, 379--389

    work page 2026

  2. [2]

    Bammey, Q. 2024. Synthbuster: Towards Detection of Diffusion Model Generated Images. IEEE Open Journal of Signal Processing, 5: 1--9

    work page 2024

  3. [3]

    Cai, Q.; Wu, C.; Zhang, Y.; Yu, J.; and Tian, X. 2025. Towards Generalizable Detector for Generated Image. In Belgrave, D.; Zhang, C.; Montoya, L. N.; Lin, H.; Pascanu, R.; Koniusz, P.; Ghassemi, M.; Chen, N.; Ru \' z, I. V. M.; and Loaiza - Bonilla, A., eds., Advances in Neural Information Processing Systems 38: Annual Conference on Neural Information Pr...

    work page 2025

  4. [4]

    Cavia, B.; Horwitz, E.; Reiss, T.; and Hoshen, Y. 2024. Real-Time Deepfake Detection in the Real-World. CoRR, abs/2406.09398

    arXiv 2024

  5. [5]

    Chen, B.; Zeng, J.; Yang, J.; and Yang, R. 2024. DRCT: Diffusion Reconstruction Contrastive Training towards Universal Detection of Diffusion Generated Images. In Salakhutdinov, R.; Kolter, Z.; Heller, K. A.; Weller, A.; Oliver, N.; Scarlett, J.; and Berkenkamp, F., eds., Forty-first International Conference on Machine Learning, ICML 2024, Vienna, Austria...

    work page 2024

  6. [6]

    Chen, R.; Xi, J.; Yan, Z.; Zhang, K.; Wu, S.; Xie, J.; Chen, X.; Xu, L.; Guan, I.; Yao, T.; and Ding, S. 2025. Dual Data Alignment Makes AI-Generated Image Detector Easier Generalizable. In Belgrave, D.; Zhang, C.; Montoya, L. N.; Lin, H.; Pascanu, R.; Koniusz, P.; Ghassemi, M.; Chen, N.; Ru \' z, I. V. M.; and Loaiza - Bonilla, A., eds., Advances in Neur...

    work page 2025

  7. [7]

    Cheng, S.; Lyu, L.; Wang, Z.; Zhang, X.; and Sehwag, V. 2025. CO-SPY: Combining Semantic and Pixel Features to Detect Synthetic Images by AI . In IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2025, Nashville, TN, USA, June 11-15, 2025 , 13455--13465. Computer Vision Foundation / IEEE

    work page 2025

  8. [8]

    Cozzolino, D.; Poggi, G.; Corvi, R.; Nie ner, M.; and Verdoliva, L. 2024. Raising the Bar of AI-generated Image Detection with CLIP . In IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2024 - Workshops, Seattle, WA, USA, June 17-18, 2024 , 4356--4366. IEEE

    work page 2024

Show all 30 references
  1. [9]

    Dhakal, A.; Khanal, S.; Sastry, S.; Arndt, J.; Ambrozio Dias, P.; Lunga, D.; and Jacobs, N. 2026. SimLBR: Learning to Detect Fake Images by Learning to Detect Real Images. In Computer Vision and Pattern Recognition. IEEE/CVF

    2026

  2. [10]

    Dosovitskiy, A.; Beyer, L.; Kolesnikov, A.; Weissenborn, D.; Zhai, X.; Unterthiner, T.; Dehghani, M.; Minderer, M.; Heigold, G.; Gelly, S.; Uszkoreit, J.; and Houlsby, N. 2021. An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale. In 9th International Con...

    2021

  3. [11]

    Guillaro, F.; Zingarini, G.; Usman, B.; Sud, A.; Cozzolino, D.; and Verdoliva, L. 2025. A Bias-Free Training Paradigm for More General AI-generated Image Detection. In IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2025, Nashville, TN, USA, June 11-15, 20...

    2025

  4. [12]

    G.; Zhu, M.; Chen, B.; Kalenichenko, D.; Wang, W.; Weyand, T.; Andreetto, M.; and Adam, H

    Howard, A. G.; Zhu, M.; Chen, B.; Kalenichenko, D.; Wang, W.; Weyand, T.; Andreetto, M.; and Adam, H. 2017. MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications. CoRR, abs/1704.04861

    2017 arXiv

  5. [13]

    Li, O.; Cai, J.; Hao, Y.; Jiang, X.; Hu, Y.; and Feng, F. 2025. Improving Synthetic Image Detection Towards Generalization: An Image Transformation Perspective. In Sun, Y.; Chierichetti, F.; Lauw, H. W.; Perlich, C.; Tok, W. H.; and Tomkins, A., eds., Proceedings of the 31st A...

    2025

  6. [14]

    Liu, H.; Tan, Z.; Tan, C.; Wei, Y.; Wang, J.; and Zhao, Y. 2024. Forgery-aware Adaptive Transformer for Generalizable Synthetic Image Detection. In IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2024, Seattle, WA, USA, June 16-22, 2024 , 10770--10780. IEEE

    2024

  7. [15]

    Liu, R.; Cui, M.; Qin, Z.; Yan, Z.; Chen, R.; Han, Y.; Li, Z.; Chen, J.; Chen, Z.; Lin, K.; Shen, J.; Weng, L.; Dong, J.; Wang, Y.; and Wu, S. 2026. MIRROR: Manifold Ideal Reference ReconstructOR for Generalizable AI-Generated Image Detection. CoRR, abs/2602.02222

    2026

  8. [16]

    Ojha, U.; Li, Y.; and Lee, Y. J. 2023. Towards Universal Fake Image Detectors that Generalize Across Generative Models. In IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2023, Vancouver, BC, Canada, June 17-24, 2023 , 24480--24489. IEEE

    2023

  9. [17]

    Oquab, M.; Darcet, T.; Moutakanni, T.; Vo, H. V.; Szafraniec, M.; Khalidov, V.; Fernandez, P.; Haziza, D.; Massa, F.; El - Nouby, A.; Assran, M.; Ballas, N.; Galuba, W.; Howes, R.; Huang, P.; Li, S.; Misra, I.; Rabbat, M.; Sharma, V.; Synnaeve, G.; Xu, H.; J \' e gou, H.; Mair...

    2024

  10. [18]

    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...

    2021

  11. [19]

    S.; Ojha, U.; Schloesser, J.; and Lee, Y

    Rajan, A. S.; Ojha, U.; Schloesser, J.; and Lee, Y. J. 2025. Aligned Datasets Improve Detection of Latent Diffusion-Generated Images. In The Thirteenth International Conference on Learning Representations, ICLR 2025, Singapore, April 24-28, 2025 . OpenReview.net

    2025

  12. [20]

    Ren, S.; Zhou, Y.; Shen, X.; Zewde, K.; Duong, T.; Huang, G.; Wei, E.; Xue, J.; et al. 2026. How well are open sourced AI-generated image detection models out-of-the-box: A comprehensive benchmark study. arXiv preprint arXiv:2602.07814

    2026

  13. [21]

    V.; Seitzer, M.; Baldassarre, F.; Oquab, M.; Jose, C.; Khalidov, V.; Szafraniec, M.; Yi, S

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

    2026

  14. [22]

    Tan, C.; Liu, H.; Zhao, Y.; Wei, S.; Gu, G.; Liu, P.; and Wei, Y. 2024. Rethinking the Up-Sampling Operations in CNN-Based Generative Network for Generalizable Deepfake Detection. In IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2024, Seattle, WA, USA, J...

    2024

  15. [23]

    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 Walsh, T.; Shah, J.; and Kolter, Z., eds., Thirty-Ninth AAAI Conference on Artificial Intelligence, Thi...

    2025

  16. [24]

    Wang, G.; Li, J.; Sun, Y.; Chen, X.; Liu, C.; Wu, Y.; Lu, M.; Song, S.; and Yadkori, Y. A. 2025. Hierarchical reasoning model. arXiv preprint arXiv:2506.21734

    2025 arXiv

  17. [25]

    Wang, S.; Wang, O.; Zhang, R.; Owens, A.; and Efros, A. A. 2020. CNN-Generated Images Are Surprisingly Easy to Spot... for Now. In 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2020, Seattle, WA, USA, June 13-19, 2020 , 8692--8701. Computer Vision F...

    2020

  18. [26]

    Wang, Z.; Bao, J.; Zhou, W.; Wang, W.; Hu, H.; Chen, H.; and Li, H. 2023. DIRE for Diffusion-Generated Image Detection. In IEEE/CVF International Conference on Computer Vision, ICCV 2023, Paris, France, October 1-6, 2023 , 22388--22398. IEEE

    2023

  19. [27]

    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 The Thirteenth International Conference on Learning Representations, ICLR 2025, Singapore, April 24-28, 2025 . OpenReview.net

    2025

  20. [28]

    Zheng, C.; Lin, C.; Zhao, Z.; Wang, H.; Guo, X.; Liu, S.; and Shen, C. 2024. Breaking Semantic Artifacts for Generalized AI-generated Image Detection. In Globersons, A.; Mackey, L.; Belgrave, D.; Fan, A.; Paquet, U.; Tomczak, J. M.; and Zhang, C., eds., Advances in Neural Info...

    2024

  21. [29]

    Zhong, N.; Xu, Y.; Li, S.; Qian, Z.; and Zhang, X. 2023. Patchcraft: Exploring Texture Patch for Efficient AI-generated Image Detection. arXiv preprint arXiv:2311.12397

    2023 arXiv

  22. [30]

    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 Oh, A.; Naumann, T.; Globerson, A.; Saenko, K.; Hardt, M.; and Levine, S., eds., Advances in Neural Infor...

    2023

Pith tools

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