Pith. sign in

REVIEW 4 major objections 6 minor 74 references

FodFoM: Fake Outlier Data by Foundation Models Creates Stronger Visual Out-of-Distribution Detector

T0 review · 4 major / 6 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read FodFoM shows that training image classifiers on fake outlier images synthesized by foundation models—semantically similar to but different from in-distribution images—produces stronger out-of-distribution detection, reaching…

desk verdict Practical fake-OOD training pipeline with strong ablations, but the core outward-shift mechanism is never verified on the generated images. read the letter →

arxiv 2412.05293 v1 pith:LSU574PC submitted 2024-11-22 cs.CV cs.LG

Jiankang Chen , Ling Deng , Zhiyong Gan , Wei-Shi Zheng , Ruixuan Wang This is my paper · ORCID
classification cs.CVcs.LG
keywords Out-of-DistributionDetectionFoundationModelsFakeOODImageGenerationStableDiffusionCLIPTextEmbeddingsBLIP-2GroundingDINOSupervisedContrastiveLearning
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 establish that a classifier's decision boundary between in-distribution and out-of-distribution images can be sharpened by training on synthetic 'fake outlier' images that are deliberately close to the in-distribution classes. It builds these images by chaining BLIP-2 captions, CLIP text embeddings, and Stable Diffusion to synthesize images that differ slightly from each ID class, plus GroundingDINO-based blurring to turn ID images into background-only versions. When these fake outliers are added to training as an extra class, with supervised contrastive loss, the model becomes less overconfident and detects real OOD images better. The paper reports new state-of-the-art results on CIFAR10, CIFAR100 and ImageNet100 benchmarks, and shows the gains hold across backbones and combine with post-hoc OOD scores.

What carries the argument

The central object is the fake OOD text embedding construction of Eq. (3), which takes a CLIP text embedding at the periphery of an ID class cluster and moves a small step gamma outward along the unit vector from the class mean; the resulting embedding conditions Stable Diffusion to generate an image that stays visually close to the ID class while falling outside it. A second mechanism uses GroundingDINO to detect foreground ID objects and blurs those regions to produce background-only fake OOD images. Together these define an extra (C+1)-th class during training, separated from ID classes by cross-entropy loss and supervised contrastive loss, and at inference only the original C logits feed the Energy-based OOD score.

What would settle it

Feed the Stable Diffusion generated fake OOD images into a strong OOD detector trained only on ID data (e.g., ViM or Energy with ReAct on the same backbone). If most fake images are scored as firmly ID (high energy score) or as arbitrary noise (extreme low density), the construction failed to produce near-boundary outliers. A second check is a sweep of the step size gamma: if performance does not peak in the middle and degrade at very small and very large gamma, the boundary-nudging story is not what drives the gain.

Watch

Extended reading notes

Core claim

The central claim is that classifiers trained with the constructed fake OOD images learn better decision boundaries for effective OOD detection. Concretely, on CIFAR10 with ResNet18 the average FPR95 drops to 8.43% and AUROC rises to 98.33%, versus 14.64% and 97.10% for the strongest baseline CIDER; on ImageNet100 with ResNet50 the average FPR95 is 33.44% and AUROC 93.79%, versus 36.46% and 92.11% for DICE. The benefit is attributed to the fact that the fake outliers are semantically similar to ID images (via the CLIP text-space perturbation) and share background context with ID (via GroundingDINO blurring), so the model is forced to separate ID from near-miss OOD cases.

Load-bearing premise

The load-bearing premise is that taking a CLIP text embedding near the edge of an ID class cluster and nudging it outward a small step produces a text prompt whose Stable Diffusion image is both visually plausible and genuinely outside the ID class; the paper does not quantitatively verify the generated images with an independent OOD score, and the step size is tuned per benchmark without sensitivity analysis.

Editorial extensions

If this is right

  • Classifiers trained with the generated fake OOD images can serve as stronger OOD detectors for the same backbone and ID dataset, improving average FPR95 and AUROC over existing training-based and post-hoc methods.
  • Existing post-hoc scores (MSP, Energy, ViM, ReAct) all improve when applied on the FodFoM-trained backbone, so the training recipe is score-agnostic.
  • The method generalizes to clean, challenging OOD benchmarks such as NINCO and OpenImage-O, where conventional evaluation sets are known to be contaminated.
  • The framework needs no external real OOD data and, per the comparison with Dream-OOD, reaches better average performance with fewer generated images.
  • The benefits persist across ResNet18/34/50/101 backbones, indicating the effect is not tied to one architecture.

Reading between the lines

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

  • One implicit implication is that the CLIP text-embedding perturbation of Eq. (3) is a transferable device: the same outward-step construction could be applied to any CLIP-based OOD detector without image generation, e.g., as negative prompts for zero-shot detection.
  • The background-blur component suggests a cheap, model-agnostic augmentation for any OOD method: any detector of foreground objects could replace GroundingDINO, making the recipe lighter.
  • Because the pipeline depends on the generative fidelity of Stable Diffusion and the alignment of CLIP text space, its quality will track improvements in those foundation models; the OOD gains are likely to grow as the generators improve.
  • A testable extension would be to use the same framework on fine-grained ID classes or medical images, where the 'semantically similar but different' construction may be more or less effective depending on how separable the classes are in CLIP space.
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 / 6 minor

Summary. The paper proposes FodFoM, a training-based out-of-distribution (OOD) detection framework that generates two types of fake OOD images using foundation models. The first type is produced by BLIP-2 image captioning, CLIP text-embedding manipulation, and Stable Diffusion generation: captions of in-distribution (ID) images are embedded with CLIP, peripheral text embeddings are shifted outward from the class-mean embedding by a step gamma (Eq. 3), and the resulting fake text embeddings condition Stable Diffusion to synthesize images that are semantically close to but (by assumption) outside the ID classes. The second type is constructed by using GroundingDINO to detect foreground objects and blurring the detected regions to create pure-background images. The classifier is trained with an extra class for the fake OOD images using a cross-entropy loss and a supervised contrastive loss. Experiments on CIFAR10, CIFAR100, and ImageNet100 with ResNet18/34/50/101 report state-of-the-art FPR95/AUROC on most benchmarks and per-OOD-dataset comparisons, along with ablations, sensitivity studies for some hyperparameters, and a comparison with Dream-OOD in the supplementary material. The central claim is that classifiers trained with these fake OOD images learn better decision boundaries for OOD detection.

Significance. If the reported results are reproducible, the paper is a valuable practical contribution: it shows that off-the-shelf foundation models can be combined to generate challenging OOD training data without GAN training instability or restrictive feature-space assumptions. The evaluation is broad, covering three ID benchmarks, multiple backbones, six OOD test sets for CIFAR, clean OOD benchmarks (NINCO, OpenImage-O), a Dream-OOD comparison, and generalization studies for the captioning model and similarity metric. The authors also release code, and the framework can be fused with several post-hoc scores, which strengthens the paper's applicability. The main weakness is that the core generative mechanism—the outward text-embedding shift—is not quantitatively verified at the image level, and the absence of error bars leaves some state-of-the-art claims under-supported.

major comments (4)
  1. [2.2.2, Eq. (3)] The paper claims that the shifted text embeddings produce 'challenging fake OOD images,' but it never quantitatively verifies that the generated images are OOD relative to the ID classes. Figure 5 shows only a few qualitative examples, and Figure 3 reports cosine similarities in the text-embedding space, which is the construction space rather than the image space. I request that the authors measure, for example, the feature distance between generated images and ID images using a pre-trained encoder, or apply an off-the-shelf OOD detector (e.g., one of the baselines from Table 1) to the generated images and report the fraction classified as OOD. Without such evidence, the reported gains cannot be attributed to the outward-shift mechanism.
  2. [3.1 and Eq. (3)] The step length gamma is set to three or five different values per benchmark, but there is no sensitivity analysis for gamma and no stated criterion for choosing these values. Since the outward shift is the core mechanism that moves text embeddings outside the ID cluster, the choice of gamma is load-bearing: too small a gamma keeps the embedding inside the cluster (so the generated images are still ID-like), and too large a gamma may produce unrelated images. Please provide a sensitivity study over gamma and describe how the values were selected (e.g., on a validation OOD set). If a validation set was used, this should be disclosed to rule out test-set overfitting.
  3. [Tables 1–3 and 5] No error bars or multiple-seed results are reported anywhere. Some of the claimed improvements over the strongest baselines are small (e.g., Table 2, ImageNet100 ResNet50 FPR95 33.44 vs. DICE+ReAct 34.75; Table 3, OpenImage-O FPR95 48.68 vs. LINe 62.15 is larger). To support the 'state-of-the-art' claim, the authors should report mean and standard deviation over at least three independent training runs for the main benchmarks, and ideally a significance test for the key comparisons.
  4. [3.3, Table 4] The ablation demonstrates that adding the SD-OOD images improves performance, but it does not isolate the effect of the outward shift itself. A control with gamma=0 (i.e., using the periphery text embeddings without any outward step) should be included; if the performance gain disappears or reverses with gamma=0, this would confirm the role of the shift, whereas if the gain persists, the improvement may be due simply to adding extra synthetic images to the (C+1)-th class. Similarly, a control with random text embeddings would clarify whether semantic proximity to the ID classes matters.
minor comments (6)
  1. [3.1] Typo: 'emebdding' should be 'embedding' in the description of the hyperparameter gamma.
  2. [3.3] Typo: 'Stabel Diffusion' should be 'Stable Diffusion'.
  3. [5] Typo: 'achived' should be 'achieved' in the conclusion.
  4. [Supplementary, Figure 6 caption] Typo: 'Comparsion' should be 'Comparison'.
  5. [2.4] Typo: 'the the class of fake OOD images' should be 'the class of fake OOD images'.
  6. [3.2.1] Grammar: 'All results supports' should be 'All results support'.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the pipeline generates synthetic training outliers from ID data and evaluates on held-out OOD benchmarks, with no fitted quantity fed back into the detection score.

full rationale

FodFoM is a training-data-augmentation method, not a derivation that smuggles its conclusion into its inputs. BLIP-2 captions ID images (Eq. 1), CLIP text embeddings are shifted outward by the paper's own Eq. (3), Stable Diffusion generates so-called fake OOD images, and GroundingDINO-plus-blur produces background images; the classifier is then trained with an extra OOD class using L_CE and L_SC (Eqs. 4-5). Evaluation uses ReAct/energy scores on held-out OOD datasets (SVHN, LSUN, iSUN, Textures, Places365, NINCO, OpenImage-O) with categories disjoint from ID. No test-set statistic is used to set gamma, alpha, beta, or any training parameter, so the reported FPR95/AUROC improvements are out-of-sample measurements rather than constructional identities. The paper's cited motivations are external works, and no uniqueness theorem or load-bearing self-citation is invoked. The genuine weakness, stated implicitly in Sec. 2.2.3 and Fig. 5, is the unverified assumption that a small CLIP text-space step (gamma between 1e-5 and 1.5e-4) produces images that are semantically OOD rather than on-manifold ID; Sec. 2.2.3 asserts: 'Since fake OOD text embeddings are semantically similar to but different from text embeddings of ID images, the generated images from Stable Diffusion with the fake OOD text embeddings as conditions would also be similar to but different from ID images.' That is a validity risk about the mechanism, not circularity, because the reported benchmark numbers would remain independent even if the mechanism failed. Overall circularity score: 0.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

The method's central mechanism depends on CLIP text-space geometry, Stable Diffusion's text-to-image fidelity, and GroundingDINO's open-set detection. These are strong but unverified domain assumptions. The hyperparameters alpha, gamma, beta, tau, lambda, and blur kernel size are hand-chosen per benchmark and contribute to the empirical result; gamma is not covered by the sensitivity study.

free parameters (6)
  • alpha (periphery selection percentage) = 30 (CIFAR), 20 (ImageNet100)
    Selects the fraction of ID text embeddings used as seed points; threshold s_alpha is derived from it. Chosen per benchmark, no validation split described.
  • gamma (outward step length) = 1e-5, 5e-5, 9e-5 (ImageNet100); 3e-5, 6e-5, 9e-5, 1.2e-4, 1.5e-4 (CIFAR)
    Controls how far fake OOD text embeddings move outside the class cluster; no sensitivity study is reported for gamma.
  • beta (foreground area threshold) = 50%
    Determines which ID images are used to create background images; images with large foregrounds are excluded.
  • tau (contrastive temperature) = 0.1
    Temperature in the supervised contrastive loss; sensitivity is shown in Figure 4.
  • lambda (loss coefficient) = 1
    Balances cross-entropy and SupCon losses; sensitivity is shown in Figure 4.
  • Mean filter kernel size = 50
    Kernel size for blurring foreground regions in background image generation.
assumptions (4)
  • domain assumption CLIP text embeddings of ID images are clustered per class in the textual semantic space.
    Sec 2.2.2: 'text embeddings of ID images are often clustered for each ID class' is the premise for using cosine similarity to the mean to identify periphery embeddings.
  • domain assumption Stable Diffusion can generate images consistent with arbitrary CLIP text embeddings, including shifted ones.
    Sec 2.2.3 feeds constructed fake OOD text embeddings directly as conditions; this requires the diffusion model to render meaningful images for out-of-cluster prompts.
  • domain assumption GroundingDINO reliably detects foreground objects of ID classes given the class name as a text query.
    Sec 2.3: foreground bounding boxes are used to blur ID objects; failure to detect leaves ID objects in the background images, making them not OOD.
  • domain assumption The OOD test sets are disjoint from ID classes and clean.
    Sec 3.1 states categories are disjoint; Sec 3.2.2 relies on NINCO and OpenImage-O being clean per Bitterwolf et al. This is taken from prior work, not re-verified.

how reviews work

0 comments
Cite this review

Pith. "Pith review of FodFoM: Fake Outlier Data by Foundation Models Creates Stronger Visual Out-of-Distribution Detector." pith.science (2026). https://pith.science/paper/LSU574PC

@misc{pith2026241205293,
  author       = {Pith},
  title        = {Pith review of: FodFoM: Fake Outlier Data by Foundation Models Creates Stronger Visual Out-of-Distribution Detector},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/LSU574PC}},
  note         = {Machine review of arXiv:2412.05293}
}
read the original abstract

Out-of-Distribution (OOD) detection is crucial when deploying machine learning models in open-world applications. The core challenge in OOD detection is mitigating the model's overconfidence on OOD data. While recent methods using auxiliary outlier datasets or synthesizing outlier features have shown promising OOD detection performance, they are limited due to costly data collection or simplified assumptions. In this paper, we propose a novel OOD detection framework FodFoM that innovatively combines multiple foundation models to generate two types of challenging fake outlier images for classifier training. The first type is based on BLIP-2's image captioning capability, CLIP's vision-language knowledge, and Stable Diffusion's image generation ability. Jointly utilizing these foundation models constructs fake outlier images which are semantically similar to but different from in-distribution (ID) images. For the second type, GroundingDINO's object detection ability is utilized to help construct pure background images by blurring foreground ID objects in ID images. The proposed framework can be flexibly combined with multiple existing OOD detection methods. Extensive empirical evaluations show that image classifiers with the help of constructed fake images can more accurately differentiate real OOD images from ID ones. New state-of-the-art OOD detection performance is achieved on multiple benchmarks. The code is available at \url{https://github.com/Cverchen/ACMMM2024-FodFoM}.

Figures

Figures reproduced from arXiv: 2412.05293 by the authors.

Figure 1
Figure 1. OOD detection performance of different methods on [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Illustration of the proposed FodFoM framework. FodFoM generates fake OOD images in two ways. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Generation of fake OOD text embeddings. Left: [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Sensitivity study of hyper-parameters 𝜆, 𝜏, and 𝛼. Dashed line: performance of the best baseline. Backbone: ResNet18 [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: Demonstrative fake OOD images generated from [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 7
Figure 7. Figure 7: Visualizations of background images generated by [PITH_FULL_IMAGE:figures/full_fig_p013_7.png]
Figure 6
Figure 6. Figure 6: Comparsion of MSP, MSP (Pre-C) and MSP (Pre [PITH_FULL_IMAGE:figures/full_fig_p013_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

74 extracted references · 58 canonical work pages

  1. [1]

    Yong Hyun Ahn, Gyeong-Moon Park, and Seong Tae Kim. 2023. LINe: Out-of- Distribution Detection by Leveraging Important Neurons. InCVPR. 19852–19862

    work page 2023

  2. [2]

    Jon Almazán, Byungsoo Ko, Geonmo Gu, Diane Larlus, and Yannis Kalantidis

    work page

  3. [3]

    Julian Bitterwolf, Maximilian Mueller, and Matthias Hein. 2023. In or Out? Fixing ImageNet Out-of-Distribution Detection Evaluation. In ICML. 2471–2506

    work page 2023

  4. [4]

    Rishi Bommasani, Drew A Hudson, Ehsan Adeli, Russ Altman, Simran Arora, Sydney von Arx, Michael S Bernstein, Jeannette Bohg, Antoine Bosselut, Emma Brunskill, et al . 2021. On the opportunities and risks of foundation models. arXiv:2108.07258

    arXiv 2021

  5. [5]

    Mathilde Caron, Hugo Touvron, Ishan Misra, Hervé Jégou, Julien Mairal, Piotr Bojanowski, and Armand Joulin. 2021. Emerging properties in self-supervised vision transformers. In ICCV. 9650–9660

    work page 2021

  6. [6]

    Jiefeng Chen, Yixuan Li, Xi Wu, Yingyu Liang, and Somesh Jha. 2021. ATOM: Robustifying Out-of-Distribution Detection Using Outlier Mining. InECML PKDD. 430–445

    work page 2021

  7. [7]

    Zhao Chen, Vijay Badrinarayanan, Chen-Yu Lee, and Andrew Rabinovich. 2018. Gradnorm: Gradient normalization for adaptive loss balancing in deep multitask networks. In ICML. 794–803

    work page 2018

  8. [8]

    Mircea Cimpoi, Subhransu Maji, Iasonas Kokkinos, Sammy Mohamed, and An- drea Vedaldi. 2014. Describing textures in the wild. In CVPR. 3606–3613

    work page 2014

Show all 74 references
  1. [9]

    Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. 2009. Imagenet: A large-scale hierarchical image database. In CVPR. 248–255

    2009

  2. [10]

    Andrija Djurisic, Nebojsa Bozanic, Arjun Ashok, and Rosanne Liu. 2022. Ex- tremely Simple Activation Shaping for Out-of-Distribution Detection. In ICLR

    2022

  3. [11]

    Xuefeng Du, Yiyou Sun, Jerry Zhu, and Yixuan Li. 2023. Dream the Impossible: Outlier Imagination with Diffusion Models. In NeurIPS

    2023

  4. [12]

    Xuefeng Du, Zhaoning Wang, Mu Cai, and Yixuan Li. 2021. VOS: Learning What You Don’t Know by Virtual Outlier Synthesis. In ICLR

    2021

  5. [13]

    Sepideh Esmaeilpour, Bing Liu, Eric Robertson, and Lei Shu. 2022. Zero-shot out-of-distribution detection based on the pre-trained model clip. In AAAI. 6568– 6576

    2022

  6. [14]

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

    2016

  7. [15]

    Dan Hendrycks, Steven Basart, Mantas Mazeika, Andy Zou, Joseph Kwon, Mo- hammadreza Mostajabi, Jacob Steinhardt, and Dawn Song. 2022. Scaling Out-of- Distribution Detection for Real-World Settings. In ICML. 8759–8773

    2022

  8. [16]

    Dan Hendrycks and Kevin Gimpel. 2016. A Baseline for Detecting Misclassified and Out-of-Distribution Examples in Neural Networks. In ICLR

    2016

  9. [17]

    Dan Hendrycks, Mantas Mazeika, and Thomas Dietterich. 2019. Deep anomaly detection with outlier exposure. In ICLR

    2019

  10. [18]

    Yen-Chang Hsu, Yilin Shen, Hongxia Jin, and Zsolt Kira. 2020. Generalized odin: Detecting out-of-distribution image without learning from out-of-distribution data. In CVPR. 10951–10960

    2020

  11. [19]

    Kushal Kafle and Christopher Kanan. 2017. Visual question answering: Datasets, algorithms, and future challenges. In CVIU. 3–20

    2017

  12. [20]

    Julian Katz-Samuels, Julia B Nakhleh, Robert Nowak, and Yixuan Li. 2022. Train- ing ood detectors in their natural habitats. In ICML. 10848–10865

    2022

  13. [21]

    Prannay Khosla, Piotr Teterwak, Chen Wang, Aaron Sarna, Yonglong Tian, Phillip Isola, Aaron Maschinot, Ce Liu, and Dilip Krishnan. 2020. Supervised contrastive learning. In NeurIPS. 18661–18673

    2020

  14. [22]

    Jaeyoung Kim, Seo Taek Kong, Dongbin Na, and Kyu-Hwan Jung. 2023. Key Fea- ture Replacement of In-Distribution Samples for Out-of-Distribution Detection. AAAI (2023)

    2023

  15. [23]

    Shu Kong and Deva Ramanan. 2021. Opengan: Open-set recognition via open data generation. In ICCV. 813–822

    2021

  16. [24]

    Krizhevsky and G

    A. Krizhevsky and G. Hinton. 2009. Learning multiple layers of features from tiny images. Technical Report. University of Toronto

    2009

  17. [25]

    Johnson Kuan and Jonas Mueller. 2022. Back to the basics: Revisiting out-of- distribution detection baselines. arXiv:2207.03061

    2022 arXiv

  18. [26]

    Kimin Lee, Kibok Lee, Honglak Lee, and Jinwoo Shin. 2018. A simple unified framework for detecting out-of-distribution samples and adversarial attacks. In NeurIPS. 7167–7177

    2018

  19. [27]

    Junnan Li, Dongxu Li, Silvio Savarese, and Steven Hoi. 2023. Blip-2: Bootstrapping language-image pre-training with frozen image encoders and large language models. arXiv:2301.12597

    2023 arXiv

  20. [28]

    Shiyu Liang, Yixuan Li, and R Srikant. 2018. Enhancing The Reliability of Out-of- distribution Image Detection in Neural Networks. In ICLR

    2018

  21. [29]

    Haotian Liu, Chunyuan Li, Qingyang Wu, and Yong Jae Lee. 2023. Visual instruc- tion tuning. arXiv:2304.08485

    2023 arXiv

  22. [30]

    Shilong Liu, Zhaoyang Zeng, Tianhe Ren, Feng Li, Hao Zhang, Jie Yang, Chunyuan Li, Jianwei Yang, Hang Su, Jun Zhu, et al. 2023. Grounding dino: Marrying dino with grounded pre-training for open-set object detection. arXiv:2303.05499

    2023 arXiv

  23. [31]

    Weitang Liu, Xiaoyun Wang, John Owens, and Yixuan Li. 2020. Energy-based out-of-distribution detection. In NeurIPS. 21464–21475

    2020

  24. [32]

    Ze Liu, Yutong Lin, Yue Cao, Han Hu, Yixuan Wei, Zheng Zhang, Stephen Lin, and Baining Guo. 2021. Swin transformer: Hierarchical vision transformer using shifted windows. In ICCV. 10012–10022

    2021

  25. [33]

    Yifei Ming, Ziyang Cai, Jiuxiang Gu, Yiyou Sun, Wei Li, and Yixuan Li. 2022. Delving into out-of-distribution detection with vision-language representations. In NeurIPS. 35087–35102

    2022

  26. [34]

    Yifei Ming, Ying Fan, and Yixuan Li. 2022. Poem: Out-of-distribution detection with posterior sampling. In ICML. 15650–15665

    2022

  27. [35]

    Yifei Ming, Yiyou Sun, Ousmane Dia, and Yixuan Li. 2023. How to Exploit Hyperspherical Embeddings for Out-of-Distribution Detection?. In ICLR

    2023

  28. [36]

    Lawrence Neal, Matthew Olson, Xiaoli Fern, Weng-Keen Wong, and Fuxin Li

  29. [37]

    Netzer, T

    Y. Netzer, T. Wang, A. Coates, A. Bissacco, and A. Y. Ng. 2011. Reading Digits in Natural Images with Unsupervised Feature Learning. In NeurIPS

    2011

  30. [38]

    Alex Nichol, Prafulla Dhariwal, Aditya Ramesh, Pranav Shyam, Pamela Mishkin, Bob McGrew, Ilya Sutskever, and Mark Chen. 2021. Glide: Towards photorealistic image generation and editing with text-guided diffusion models. arXiv:2112.10741

    2021 arXiv

  31. [39]

    OpenAI. 2022. Introducing ChatGPT. https://openai.com/blog/chatgpt. Accessed: 2023-03-15

    2022

  32. [40]

    Haoxuan Qu, Xiaofei Hui, Yujun Cai, and Jun Liu. 2023. LMC: Large Model Col- laboration with Cross-assessment for Training-Free Open-Set Object Recognition. In NeurIPS

    2023

  33. [41]

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. 2021. Learning transferable visual models from natural language supervision. In ICML. 8748–8763

    2021

  34. [42]

    Aditya Ramesh, Mikhail Pavlov, Gabriel Goh, Scott Gray, Chelsea Voss, Alec Radford, Mark Chen, and Ilya Sutskever. 2021. Zero-shot text-to-image generation. In ICML. 8821–8831

    2021

  35. [43]

    Jie Ren, Peter J Liu, Emily Fertig, Jasper Snoek, Ryan Poplin, Mark Depristo, Joshua Dillon, and Balaji Lakshminarayanan. 2019. Likelihood ratios for out-of- distribution detection. In NeurIPS

    2019

  36. [44]

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Björn Ommer. 2022. High-resolution image synthesis with latent diffusion models. In CVPR. 10684–10695

    2022

  37. [45]

    Chitwan Saharia, Jonathan Ho, William Chan, Tim Salimans, David J Fleet, and Mohammad Norouzi. 2022. Image super-resolution via iterative refinement. IEEE TPAMI 45, 4 (2022), 4713–4726

    2022

  38. [46]

    Mohammadreza Salehi, Hossein Mirzaei, Dan Hendrycks, Yixuan Li, Moham- mad Hossein Rohban, and Mohammad Sabokrou. 2021. A unified survey on anomaly, novelty, open-set, and out-of-distribution detection: Solutions and future challenges. arXiv:2110.14051

    2021 arXiv

  39. [47]

    Vikash Sehwag, Mung Chiang, and Prateek Mittal. 2021. SSD: A Unified Frame- work for Self-Supervised Outlier Detection. In ICLR

    2021

  40. [48]

    Zheyan Shen, Jiashuo Liu, Yue He, Xingxuan Zhang, Renzhe Xu, Han Yu, and Peng Cui. 2021. Towards Out-Of-Distribution Generalization: A Survey. arXiv:2108.13624

    2021 arXiv

  41. [49]

    Yang Shu, Xingzhuo Guo, Jialong Wu, Ximei Wang, Jianmin Wang, and Mingsheng Long. 2023. CLIPood: Generalizing CLIP to Out-of-Distributions. arXiv:2302.00864

    2023 arXiv

  42. [50]

    Matteo Stefanini, Marcella Cornia, Lorenzo Baraldi, Silvia Cascianelli, Giuseppe Fiameni, and Rita Cucchiara. 2022. From show to tell: A survey on deep learning- based image captioning. IEEE TPAMI 45, 1 (2022), 539–559

    2022

  43. [51]

    Yiyou Sun, Chuan Guo, and Yixuan Li. 2021. React: Out-of-distribution detection with rectified activations. In NeurIPS. 144–157

    2021

  44. [52]

    Yiyou Sun and Yixuan Li. 2022. DICE: Leveraging sparsification for out-of- distribution detection. In ECCV. 691–708

    2022

  45. [53]

    Yiyou Sun, Yifei Ming, Xiaojin Zhu, and Yixuan Li. 2022. Out-of-distribution detection with deep nearest neighbors. In ICML. 20827–20840

    2022

  46. [54]

    Jihoon Tack, Sangwoo Mo, Jongheon Jeong, and Jinwoo Shin. 2020. Csi: Novelty detection via contrastive learning on distributionally shifted instances. InNeurIPS. 11839–11852

    2020

  47. [55]

    Leitian Tao, Xuefeng Du, Jerry Zhu, and Yixuan Li. 2023. Non-parametric Outlier Synthesis. In ICLR

    2023

  48. [56]

    Grant Van Horn, Oisin Mac Aodha, Yang Song, Yin Cui, Chen Sun, Alex Shepard, Hartwig Adam, Pietro Perona, and Serge Belongie. 2018. The inaturalist species classification and detection dataset. In CVPR. 8769–8778

    2018

  49. [57]

    Hualiang Wang, Yi Li, Huifeng Yao, and Xiaomeng Li. 2023. Clipn for zero-shot ood detection: Teaching clip to say no. In ICCV. 1802–1812. MM ’24, October 28-November 1, 2024, Melbourne, VIC, Australia Jiankang Chen, Ling Deng, Zhiyong Gan, Wei-Shi Zheng, Ruixuan Wang

    2023

  50. [58]

    Haoqi Wang, Zhizhong Li, Litong Feng, and Wayne Zhang. 2022. Vim: Out-of- distribution with virtual-logit matching. In CVPR. 4921–4930

    2022

  51. [59]

    Hongxin Wei, Renchunzi Xie, Hao Cheng, Lei Feng, Bo An, and Yixuan Li. 2022. Mitigating neural network overconfidence with logit normalization. In ICML. 23631–23644

    2022

  52. [60]

    Jianxiong Xiao, James Hays, Krista A Ehinger, Aude Oliva, and Antonio Torralba

  53. [61]

    Pingmei Xu, Krista A Ehinger, Yinda Zhang, Adam Finkelstein, Sanjeev R Kulka- rni, and Jianxiong Xiao. 2015. Turkergaze: Crowdsourcing saliency with webcam based eye tracking. arXiv:1504.06755

    2015 arXiv

  54. [62]

    Fisher Yu, Ari Seff, Yinda Zhang, Shuran Song, Thomas Funkhouser, and Jianxiong Xiao. 2015. Lsun: Construction of a large-scale image dataset using deep learning with humans in the loop. arXiv:1506.03365

    2015 arXiv

  55. [63]

    Yeonguk Yu, Sungho Shin, Seongju Lee, Changhyun Jun, and Kyoobin Lee. 2023. Block Selection Method for Using Feature Norm in Out-of-Distribution Detection. In CVPR. 15701–15711

    2023

  56. [64]

    Jinsong Zhang, Qiang Fu, Xu Chen, Lun Du, Zelin Li, Gang Wang, Shi Han, Dong- mei Zhang, et al. 2022. Out-of-Distribution Detection based on In-Distribution Data Patterns Memorization with Modern Hopfield Energy. In ICLR

    2022

  57. [65]

    Renrui Zhang, Xiangfei Hu, Bohao Li, Siyuan Huang, Hanqiu Deng, Yu Qiao, Peng Gao, and Hongsheng Li. 2023. Prompt, generate, then cache: Cascade of foundation models makes strong few-shot learners. In CVPR. 15211–15222

    2023

  58. [66]

    Renrui Zhang, Wei Zhang, Rongyao Fang, Peng Gao, Kunchang Li, Jifeng Dai, Yu Qiao, and Hongsheng Li. 2022. Tip-adapter: Training-free adaption of clip for few-shot classification. In ECCV. 493–510

    2022

  59. [67]

    Susan Zhang, Stephen Roller, Naman Goyal, Mikel Artetxe, Moya Chen, Shuohui Chen, Christopher Dewan, Mona Diab, Xian Li, Xi Victoria Lin, et al. 2022. Opt: Open pre-trained transformer language models. (2022). arXiv:2205.01068

    2022 arXiv

  60. [68]

    Bolei Zhou, Agata Lapedriza, Aditya Khosla, Aude Oliva, and Antonio Torralba

  61. [69]

    Yao Zhu, YueFeng Chen, Chuanlong Xie, Xiaodan Li, Rong Zhang, Hui Xue, Xiang Tian, Yaowu Chen, et al. 2022. Boosting Out-of-distribution Detection with Typical Features. In NeurIPS. 20758–20769. 6 Supplementary 6.1 Datasets details 6.1.1 CIFAR benchmarks. For the CIFAR10 and C...

    2022

  62. [74]

    The best and second-best results are indicated in bold and underline

    The Euclidean distance calculates the difference between the MM ’24, October 28-November 1, 2024, Melbourne, VIC, Australia Jiankang Chen, Ling Deng, Zhiyong Gan, Wei-Shi Zheng, Ruixuan Wang Table 7: Comparison between different methods in OOD detection on ImageNet100 Benchmar...

    2024

  63. [2010]

    Sun database: Large-scale scene recognition from abbey to zoo. In CVPR. 3485–3492

  64. [2017]

    IEEE TPAMI 40, 6 (2017), 1452–1464

    Places: A 10 million image database for scene recognition. IEEE TPAMI 40, 6 (2017), 1452–1464

    2017

  65. [2018]

    Open set learning with counterfactual images. In ECCV. 613–628

  66. [2022]

    Granularity-aware adaptation for image retrieval over multiple tasks. In ECCV. 389–406

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.