Pith. sign in

REVIEW 4 major objections 4 minor 123 references

Suppress and Diversify: Refining Robust Pathways for Corruption Robustness

T0 review · 4 major / 4 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read Deep networks progressively lose their corruption-proof features as information flows deeper, and the paper shows that keeping those features alive via selection and diversification improves robustness across tasks.

desk verdict A solid, novel empirical method with a real benchmark-selection soft spot; deserves peer review but needs revision to justify the generality claims. read the letter →

arxiv 2608.06712 v1 pith:MVLDDBAS submitted 2026-08-07 cs.CV

Jiangang Yang , Wenhui Shi , Xiaoran Xu , Wenyue Chong , Luqing Luo , Jing Xing , Jian Liu This is my paper · ORCID
classification cs.CV
keywords corruptionrobustnessrobustfeaturescomputationalpathwaysneuralnetworkanalysisdataaugmentationImageNet-Csemanticsegmentationtest-timeadaptation
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

Deep neural networks are at their most corruption-robust near the input and steadily lose that robustness in deeper layers: the pool of features that barely change under noise, fog, or blur shrinks as information propagates. The paper shows that this shrinkage is not merely descriptive—preserving the surviving robust pathways while silencing the non-robust ones directly narrows the accuracy gap between clean and corrupted inputs, and overall model robustness is functionally tied to how many robust features remain. Building on this, the paper introduces Suppress and Diversify (S&D), a training-time refinement that selects robust computational pathways with a memory-aware scoring mechanism and then diversifies them through channel shuffling, spatial flips, and transposition. The result is consistent corruption-robustness gains across classification, detection, and segmentation benchmarks, achieved without new parameters, architectural changes, or test-time cost. The authors acknowledge a clean-accuracy trade-off and that the pathway selection is driven by a single synthetic corruption.

What carries the argument

The central object is the computational pathway: the function computed by a single channel (neuron) of a layer, from input to that channel's feature map. A pathway is γ-robust for corruption h when the expected normalized distance between its clean and corrupted outputs, measured by the local Lipschitz metric $D_{\mathrm{Lip}}(z, \hat{z}) = \|z - \hat{z}\| / \|x - h(x)\|$, stays below $\gamma$. This object carries the whole argument: it quantifies the layer-wise decay of robustness (Remark 1), supports the causal isolation experiments that establish functional dependence (Remark 2), and is exactly what MDSM selects and SPTS diversifies. The selection uses a fitness score $S(G_t) = \frac{1}{K} \sum_k R(P(x), P(\hat{x})) + \lambda \sigma(n/N)$, where R is a similarity metric such as CKA.

What would settle it

Train a ResNet-50 with S&D using only low-frequency noise as the selection corruption, then evaluate on blur and weather corruptions excluded from selection; if the mean corruption error on those held-out types is unchanged relative to a plain-trained baseline, the pathway-selection transfer claim fails. Conversely, if the reported ImageNet-C gains persist under such a holdout, the functional-dependence claim is supported.

Watch

Extended reading notes

Core claim

Under natural image corruptions, robust features—defined via the local Lipschitz constant as features whose activation distance stays below a threshold when the input is corrupted—are common in early layers and decay progressively as information flows deeper, particularly under stringent thresholds. Linear probing and loss-landscape analysis show that selectively activating γ-robust pathways while zeroing non-robust ones produces a markedly smaller robustness gap at every sub-network and for the full model, and that robust pathways sit in flatter regions of the loss landscape. The paper concludes that robustness is functionally dependent on the prevalence of these features. It then operationalizes the finding in S&D: the Memory-aware Dynamic Selection Mechanism (MDSM) maintains a candidate pathway group in a memory bank, scoring each group by a fitness function that combines clean–corrupted feature similarity (CKA) with a sigmoid-scheduled discriminability term, and the Structure-consistent Path Tweaking Strategy (SPTS) retains the top-K robust pathways and regenerates diversity through channel shuffling, horizontal/vertical flips, and matrix transpose. The refined pathway group is folded into standard training, with the learning rate on the refined group set to zero to preserve the robust features.

Load-bearing premise

The whole pipeline hinges on the assumption that contrast-plus-noise, the single synthetic corruption used to score and select robust pathways during training, is representative enough of the fifteen ImageNet-C corruptions (and the other held-out corruptions) that the selected pathways remain robust beyond the selection distribution.

Editorial extensions

If this is right

  • If robust features decay with depth, then methods that address robustness only at the input (augmentation) or only at the output (regularization) are attacking the symptom, not the decay; pathway-level refinement is a complementary lever.
  • S&D claims to be orthogonal: stacking it on top of AugMix, Mixup, and modern training recipes yields additional gains, suggesting it composes with existing robustness methods rather than replacing them.
  • Zero test-time overhead means the refinement costs nothing at inference; gains come entirely from training-time pathway selection and diversification.
  • Because S&D is architecture-agnostic, the same recipe extends from CNNs to ViTs and Mamba-style models, and from classification to detection and segmentation.
  • The reported gains on test-time adaptation benchmarks indicate that models refined by S&D also provide a better starting point for adaptation methods.

Reading between the lines

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

  • The decay phenomenon suggests a testable prediction about scale: larger models, which have more capacity in deep layers, should show either a shallower decay or a larger pool of robust features; if neither holds, the claimed functional dependence may break under scaling.
  • The dependence on a single selection corruption is the transfer risk: a practitioner looking to deploy S&D should first check which synthetic corruption best matches their deployment conditions, since the paper's published gains are validated on the ImageNet-C family that may correlate with the chosen Contrast+Noise transform.
  • Symmetry-preserving diversifications (flips, transpose) are inherently spatial; the same logic would need a different diversification family for temporal or non-Euclidean features, which the paper does not address.
  • If the functional-dependence claim is right, it suggests a cheap diagnostic: monitoring the prevalence of γ-robust features in a deployed model could serve as an early warning that robustness is degrading before errors appear.
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 / 4 minor

Summary. This paper proposes a pathway-level account of corruption robustness. It defines γ-robust features and pathways via a local Lipschitz distortion metric D_Lip (Eq. 4), observes that the proportion of γ-robust features decreases from shallow to deep sub-networks (Figs. 2–3), and claims a functional dependency between robust-feature prevalence and model robustness (Remark 2, supported only by binary pathway-ablation and linear-probing experiments in §3.2.3). Based on this, it introduces S&D, consisting of MDSM (§3.3.2), which selects pathways whose CKA similarity under a synthetic Contrast+Noise corruption is high, and SPTS (§3.3.3), which keeps the top half and augments them with symmetry-preserving weight transforms. Experiments across eight benchmarks report consistent corruption-robustness gains for classification, detection, segmentation, and test-time adaptation, with small clean-accuracy losses.

Significance. If its central claims held, the paper would provide a useful characterization of internal robustness and a lightweight, task-agnostic refinement. The authors ship code, evaluate across many architectures and tasks, report training overhead, and include a per-corruption breakdown (Table D.2) and extensive ablations (Appendix B). These are real strengths. However, the headline claims ('functional dependency', 'parameter-free', consistent gains) outrun the evidence. The selection signal is tuned on the target benchmark and overlaps with several evaluated corruptions, the functional-dependency claim is qualitative, main tables lack error bars while ablation tables show the variability of the headline numbers, and 'parameter-free' is contradicted by several tuned hyperparameters. These issues are addressable but currently leave the transfer claim under-supported.

major comments (4)
  1. [§3.2.3, Remark 2] The claim that 'the robustness of the full model is functionally dependent on the prevalence of robust features' is not supported by the presented evidence. The causal analysis in §3.2.3 partitions stem-stage pathways into exactly two groups (top vs. bottom 50% of D_Lip scores) and compares robustness gaps, so it tests a binary presence/absence contrast, not a functional dependency. A quantitative dependency would require varying the retained fraction of robust pathways over several values and layers and showing a monotone relationship with the robustness gap or corrupted accuracy; the current linear-probing comparison (Figure 4) is qualitative. Please either add such an analysis or weaken the 'functionally dependent' phrasing in the abstract, Remark 2, and conclusion to 'associated with' or 'consistent with a monotone relationship'.
  2. [§3.3.2, Eq. (6), Appendix B.3, Table D.2] The pathway-selection signal in MDSM is computed with a single synthetic transform, Contrast+Noise, and Appendix B.3 shows that this transform was chosen by comparing ImageNet-C validation performance among seven candidates. The headline results are then reported on ImageNet-C and its variants, and the per-corruption breakdown in Table D.2 shows the largest improvements on corruptions that are kindred to the selection signal (contrast, brightness, several blurs). This is benchmark-informed selection of the training-time corruption signal and can inflate the transfer numbers. Please report an evaluation where the selection transform is fixed a priori (e.g., a single transform or a small set used across all benchmarks without validation-based selection), and/or show per-corruption gains on held-out corruption families that are not kindred to Contrast+Noise.
  3. [Tables 1, B.1, B.3, B.5, B.7] The main tables report single runs without error bars, and the ablation tables give different values for what appears to be the same default configuration on ResNet-18: B.1 reports ImageNet-C 34.8 with MDSM+SPTS, B.3 reports 34.9 for Contrast+Noise, B.4 reports 34.8 for CKA, and B.5 reports 34.3 for the default 50% TopK, while the seed sensitivity in B.7 ranges from 34.2 to 37.6. Without mean±std over multiple seeds in the main tables, the claimed consistent gains of 0.4–2.5% in Table 1 may fall within seed noise. Please provide multi-seed statistics, especially for Table 1, Table 3, and the key ablation tables, and explain the discrepancies among the ablation tables.
  4. [Abstract, §3.3.1, Eq. (6), Eq. (7), Appendix B.8] The paper repeatedly calls S&D 'parameter-free' (Abstract, §3.3.1, Conclusion), but several tuned hyperparameters are listed: λ and N in Eq. (6), the TopK ratio K/|G'| in Eq. (7), the corruption transform for x-hat (Appendix B.3), and the learning rate for G* (Appendix B.8). If 'parameter-free' means 'no additional trainable parameters' rather than 'no hyperparameters', please state this explicitly and list all fixed hyperparameters and their chosen values in the main text; otherwise remove the 'parameter-free' claim from the abstract and conclusion.
minor comments (4)
  1. [Section 5.2, Table 9] MDSM throughput is reported as 6003.23 img/s versus 880.79 img/s for baseline, a 6.8x increase despite the method adding forward passes and CKA computation; this appears to be a typo or a mismatch of units (e.g., images per second vs. batches per second) and should be corrected.
  2. [Section 4.2.1, Tables 1 and D.1] Per-corruption mCE values in Table D.2 exceed 100 (e.g., 115.6 for MobileNetV2* under JPEG), while Table 1 reports an average mCE of 86.1 for the same model; clarify the mCE normalization and explain how values above 100 arise.
  3. [Section 3.2.2, Figures 2–3] The notation f≤1, f≤5, f≤9 in Figures 2–3 should be defined explicitly at first use; f≤l is introduced in §3.1, but the reader must infer the correspondence to the stages of the ResNet-18 network from context.
  4. [Appendix A.2, Eq. (A.2)] The ensemble rule uses a fixed l2-norm threshold of 0.5 with no sensitivity analysis; please report the accuracy of this clean/corrupted classifier and a small sweep over thresholds to show that the ensemble results in Table 1 do not depend on a carefully chosen operating point.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity; central claims are empirically grounded despite minor definitional overlap.

full rationale

The paper's central derivation chain is empirical rather than definitional. Robust features are defined in Definitions 3.1 and 3.2 via feature stability under corruption, and the progressive decay (Remark 1) is a measured distributional fact about pretrained networks, not an implication of the definition; it could have failed and is instead confirmed per layer and per corruption. The functional-dependency claim (Remark 2) is supported by an intervention (zeroing pathway groups) with linear probing and loss-landscape analysis; although the grouping variable is the same stability metric, the outcome is clean-to-corrupted accuracy, so the link is not equationally forced. MDSM's fitness score in Eq. (6) intentionally instantiates the same stability criterion to select pathways; this is the method's design, and its validity is tested on held-out corruptions (ImageNet-3DCC, COCO-C, ACDC, and the per-corruption breakdown) not used to choose the Contrast+Noise transform. Appendix B.3's selection of Contrast+Noise on ImageNet-C validation is a benchmark-tuning risk rather than a definitional circularity, because the reported gains are not a fitted function of that choice and many gains appear on corruptions unlike the selection transform. No load-bearing self-citation, imported uniqueness theorem, or ansatz-smuggling citation appears. The sole caveat is a mild definitional overlap between the robustness metric and the selection signal, which limits the strength of the 'discovery' framing but does not reduce the main results to their inputs.

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

The central analysis and method rest on empirical choices: the robustness metric, the corruption transform used for selection, and several hyperparameters. The method introduces no new entities or trainable parameters beyond the architecture, but it is not 'parameter-free' in the sense of having no hyperparameters.

free parameters (5)
  • lambda = 1
    Balances feature stability R and the discriminability schedule in Eq. (6); chosen by hand, not learned.
  • N (MDSM steps) = 5
    Number of scheduled MDSM update steps per epoch; fixed in the implementation details.
  • TopK ratio K/|G'| = 0.5
    SPTS retains half of the selected pathways in Eq. (7); ablation in Appendix B.5.
  • Corruption transform for MDSM = Contrast+Noise
    Selected among seven options by ImageNet-C validation performance in Table B.3, so the evaluation benchmark informed the choice.
  • Learning rate for G* = 0.0
    The refined pathways are frozen during training; chosen from the sweep in Table B.8.
assumptions (5)
  • domain assumption Covariate shift assumption: P(X) differs from P(h(X)) but the label remains invariant under corruption.
    Stated in Section 3.1 and used to frame the optimization objective in Eq. (1).
  • domain assumption The local Lipschitz ratio D_Lip with normalized Frobenius norms is an adequate distortion metric for characterizing feature robustness under natural corruptions.
    Used in Eq. (4) and throughout Section 3.2.2 to define gamma-robust features and pathways.
  • domain assumption CKA similarity in Eq. (6) supports pathway selection and transfers to final robustness across corruption types.
    Adopted as the default similarity metric R in Section 3.3.2, with ablation in Appendix B.4 showing different metrics give different clean-robust trade-offs.
  • ad hoc to paper Symmetry-preserving weight transformations (horizontal flip, vertical flip, matrix transpose, channel shuffle) produce pathways that preserve the robust properties of the original group.
    Introduced in Section 3.3.3 and Appendix A.1; no proof is given that structural invariants are preserved, only empirical ablation in Table B.2.
  • domain assumption The progressive decay observation, demonstrated on pretrained ResNet-18, generalizes to other architectures and scales.
    Section 3.2.2 reports the phenomenon for ResNet-18 under specific corruptions; the paper applies S&D to many backbones without verifying the decay observation for each.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Suppress and Diversify: Refining Robust Pathways for Corruption Robustness." pith.science (2026). https://pith.science/paper/MVLDDBAS

@misc{pith2026260806712,
  author       = {Pith},
  title        = {Pith review of: Suppress and Diversify: Refining Robust Pathways for Corruption Robustness},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MVLDDBAS}},
  note         = {Machine review of arXiv:2608.06712}
}
read the original abstract

Model robustness against natural image corruptions is essential for safety-critical applications. While existing methods primarily focus on implicit representation learning, we provide the first systematic exploration of computational pathways to explicitly characterize internal robustness. We identify a progressive decay of robust features across network layers and establish a functional dependency between the prevalence of these features and model performance. To exploit these insights, we propose Suppress and Diversify (S\&D), a non-intrusive refinement approach that enhances robustness by dynamically selecting robust pathways and diversifying them through symmetry-preserving transformations. S\&D is architecture-agnostic, parameter-free, and incurs zero test-time overhead. Extensive evaluations across eight benchmarks demonstrate that S\&D consistently improves performance across multiple vision tasks, diverse backbones, and complex real-world scenarios, highlighting its broad efficacy and scalability.

Figures

Figures reproduced from arXiv: 2608.06712 by the authors.

Figure 1
Figure 1. (a) The phenomenon of robust (non-robust) features for image corruptions . (b) The performance gains of the proposed approach across various computer vision tasks. et al., 2023; Zeng et al., 2024), raising serious security con￾cerns in safety-critical applications. To solve this problem, prior approaches primarily focus on implicitly learning robust representations through data aug￾mentation (Hendrycks et al., 2019;… view at source ↗
Figure 2
Figure 2. The cumulative distribution of γ-robust features at different sub-networks under the corruption of (a) Gaussian Noise and (b) Fog. The consistent shift toward higher thresholds in deeper layers indicates a progressive scarcity of robust features as the network depth increases, revealing that high-level representations are more sensitive to structured corruptions. (a) (b) [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Robust feature proportions across thresholds γ under (a) Gaussian Noise and (b) Fog. The downward trend across stages (f≤1 → f≤9) reveals systemic robustness decay. grated into existing deep-learning frameworks. 3. Methodology 3.1. Problem Formulation We consider the robustness of a neural network fθ under natural image corruptions. Formally, let D = {(xi , yi)} N i=1 be a clean dataset from P(X, Y ). A corruption f… view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Impact of pathway robustness on sub-networks and the full model. We report the robustness gap (lower is better) for (a) f≤1, (b) f≤5, (c) f≤9, and (d) fθ when preserving either γ-robust or non-γ-robust pathways. (a) (b) [PITH_FULL_IMAGE:figures/full_fig_p004_4.png]
Figure 5
Figure 5. Figure 5: Loss landscapes of fθ under input perturbations. (a) Non-γ-robust pathways exhibit a sharp landscape, whereas (b) γ-robust pathways show a flatter surface. Szegedy et al., 2013) as the distortion metric D in Equa￾tion (2), denoted as DLip: DLip(zl,k, zˆl,k) = ||zl,k − …
Figure 6
Figure 6. Figure 6: The pipeline of (a) S&D, (b) MDSM, and (c) SPTS. NR, R and TR denote non-robust, robust and tweaked robust pathways, respectively. ensures stable predictions in the vicinity of clean inputs and effectively minimizes the robustness gap. Remark 2. The robustness of the f…
Figure 7
Figure 7. Figure 7: The detection results in night-rainy scenarios. (a) (b) (c) (d) [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]
Figure 8
Figure 8. Figure 8: Comparison of (a) CKA between the baseline and S&D, (b) CKA between AugMix and AugMix+S&D, and robustness gap with and without S&D on (c) ImageNet-C and (d) ImageNet￾3DCC. These results demonstrate S&D’s broad compatibility in complex, real-world degraded environments.…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

123 extracted references · 28 canonical work pages

  1. [1]

    arXiv preprint arXiv:1903.12261 , year=

    Benchmarking neural network robustness to common corruptions and perturbations , author=. arXiv preprint arXiv:1903.12261 , year=

    arXiv 1903

  2. [2]

    arXiv preprint arXiv:1907.07484 , year=

    Benchmarking robustness in object detection: Autonomous driving when winter is coming , author=. arXiv preprint arXiv:1907.07484 , year=

    arXiv 1907

  3. [3]

    International journal of computer vision , volume=

    Benchmarking the robustness of semantic segmentation models with respect to common corruptions , author=. International journal of computer vision , volume=. 2021 , publisher=

    2021

  4. [4]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Benchmarking robustness of 3d object detection to common corruptions , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  5. [5]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Benchmarking the Robustness of Temporal Action Detection Models Against Temporal Corruptions , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  6. [6]

    arXiv preprint arXiv:1912.02781 , year=

    Augmix: A simple data processing method to improve robustness and uncertainty , author=. arXiv preprint arXiv:1912.02781 , year=

    arXiv 1912

  7. [7]

    European Conference on Computer Vision , pages=

    Prime: A few primitives can boost robustness to common corruptions , author=. European Conference on Computer Vision , pages=. 2022 , organization=

    2022

  8. [8]

    International Conference on Learning Representations , year=

    mixup: Beyond Empirical Risk Minimization , author=. International Conference on Learning Representations , year=

Show all 123 references
  1. [9]

    Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    Improving robustness against common corruptions with frequency biased models , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

  2. [10]

    Advances in neural information processing systems , volume=

    A winning hand: Compressing deep networks can improve out-of-distribution robustness , author=. Advances in neural information processing systems , volume=

  3. [11]

    International Conference on Machine Learning , pages=

    Lcanets: Lateral competition improves robustness against corruption and attack , author=. International Conference on Machine Learning , pages=. 2022 , organization=

    2022

  4. [12]

    Advances in Neural Information Processing Systems , volume=

    Simulating a primary visual cortex at the front of CNNs improves robustness to image perturbations , author=. Advances in Neural Information Processing Systems , volume=

  5. [13]

    Advances in neural information processing systems , volume=

    Adversarial examples are not bugs, they are features , author=. Advances in neural information processing systems , volume=

  6. [14]

    Advances in Neural Information Processing Systems , volume=

    Can adversarial training be manipulated by non-robust features? , author=. Advances in Neural Information Processing Systems , volume=

  7. [15]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Feature separation and recalibration for adversarial robustness , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  8. [16]

    IEEE Transactions on Cybernetics , year=

    Exploring robust features for improving adversarial robustness , author=. IEEE Transactions on Cybernetics , year=

  9. [17]

    International Conference on Learning Representations , year=

    Evaluating the Robustness of Neural Networks: An Extreme Value Theory Approach , author=. International Conference on Learning Representations , year=

  10. [18]

    Advances in Neural Information Processing Systems , volume=

    Improving robustness to corruptions with multiplicative weight perturbations , author=. Advances in Neural Information Processing Systems , volume=

  11. [19]

    Advances in Neural Information Processing Systems , volume=

    Enhance the visual representation via discrete adversarial training , author=. Advances in Neural Information Processing Systems , volume=

  12. [20]

    International conference on machine learning , pages=

    Parseval networks: Improving robustness to adversarial examples , author=. International conference on machine learning , pages=. 2017 , organization=

    2017

  13. [21]

    2nd International Conference on Learning Representations, ICLR 2014 , year=

    Intriguing properties of neural networks , author=. 2nd International Conference on Learning Representations, ICLR 2014 , year=

    2014

  14. [22]

    Advances in neural information processing systems , volume=

    Efficient and accurate estimation of lipschitz constants for deep neural networks , author=. Advances in neural information processing systems , volume=

  15. [23]

    Advances in neural information processing systems , volume=

    Generalisation in humans and deep neural networks , author=. Advances in neural information processing systems , volume=

  16. [24]

    International Conference on Learning Representations , year=

    ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness , author=. International Conference on Learning Representations , year=

  17. [25]

    arXiv preprint arXiv:2104.09789 , year=

    Does enhanced shape bias improve neural network robustness to common corruptions? , author=. arXiv preprint arXiv:2104.09789 , year=

    arXiv

  18. [26]

    Advances in Neural Information Processing Systems , volume=

    A fourier perspective on model robustness in computer vision , author=. Advances in Neural Information Processing Systems , volume=

  19. [27]

    Advances in Neural Information Processing Systems , volume=

    Boundary thickness and robustness in learning models , author=. Advances in Neural Information Processing Systems , volume=

  20. [28]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Can Biases in ImageNet Models Explain Generalization? , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  21. [29]

    Proceedings of the IEEE/CVF international conference on computer vision , pages=

    The many faces of robustness: A critical analysis of out-of-distribution generalization , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=

  22. [30]

    International Journal of Computer Vision , volume=

    Mixstyle neural networks for domain generalization and adaptation , author=. International Journal of Computer Vision , volume=. 2024 , publisher=

    2024

  23. [31]

    Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    Hybridaugment++: Unified frequency spectra perturbations for model robustness , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

  24. [32]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Fourier-basis functions to bridge augmentation gap: Rethinking frequency augmentation in image classification , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  25. [33]

    arXiv preprint arXiv:2110.00476 , year=

    Resnet strikes back: An improved training procedure in timm , author=. arXiv preprint arXiv:2110.00476 , year=

    arXiv

  26. [34]

    International conference on machine learning , pages=

    Training data-efficient image transformers & distillation through attention , author=. International conference on machine learning , pages=. 2021 , organization=

    2021

  27. [35]

    PyTorch , year=

    How to train state-of-the-art models using torchvision’s latest primitives , author=. PyTorch , year=

  28. [36]

    Advances in neural information processing systems , volume=

    Learning from brains how to regularize machines , author=. Advances in neural information processing systems , volume=

  29. [37]

    Advances in Neural Information Processing Systems , volume=

    Towards robust vision by multi-task learning on monkey visual cortex , author=. Advances in Neural Information Processing Systems , volume=

  30. [38]

    European Conference on Computer Vision , pages=

    Improving robustness by enhancing weak subnets , author=. European Conference on Computer Vision , pages=. 2022 , organization=

    2022

  31. [39]

    International Conference on Machine Learning , pages=

    Cifs: Improving adversarial robustness of cnns via channel-wise importance-based feature selection , author=. International Conference on Machine Learning , pages=. 2021 , organization=

    2021

  32. [40]

    IEEE Transactions on Image Processing , volume=

    Interpreting and improving adversarial robustness of deep neural networks with neuron sensitivity , author=. IEEE Transactions on Image Processing , volume=. 2020 , publisher=

    2020

  33. [41]

    International Conference on Machine Learning , pages=

    Adversarial neural pruning with latent vulnerability suppression , author=. International Conference on Machine Learning , pages=. 2020 , organization=

    2020

  34. [42]

    Advances in Neural Information Processing Systems , volume=

    On interaction between augmentations and corruptions in natural corruption robustness , author=. Advances in Neural Information Processing Systems , volume=

  35. [43]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    3d common corruptions and data augmentation , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  36. [44]

    International conference on machine learning , pages=

    Do imagenet classifiers generalize to imagenet? , author=. International conference on machine learning , pages=. 2019 , organization=

    2019

  37. [45]

    Computer Vision--ECCV 2014: 13th European Conference, Zurich, Switzerland, September 6-12, 2014, Proceedings, Part V 13 , pages=

    Microsoft coco: Common objects in context , author=. Computer Vision--ECCV 2014: 13th European Conference, Zurich, Switzerland, September 6-12, 2014, Proceedings, Part V 13 , pages=. 2014 , organization=

    2014

  38. [46]

    International Journal of Computer Vision , volume=

    Semantic understanding of scenes through the ade20k dataset , author=. International Journal of Computer Vision , volume=. 2019 , publisher=

    2019

  39. [47]

    Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    The cityscapes dataset for semantic urban scene understanding , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

  40. [48]

    International journal of computer vision , volume=

    The pascal visual object classes challenge: A retrospective , author=. International journal of computer vision , volume=. 2015 , publisher=

    2015

  41. [49]

    Advances in neural information processing systems , volume=

    Imagenet classification with deep convolutional neural networks , author=. Advances in neural information processing systems , volume=

  42. [50]

    Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    Mobilenetv2: Inverted residuals and linear bottlenecks , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

  43. [51]

    arXiv preprint arXiv:1602.07360 , year=

    SqueezeNet: AlexNet-level accuracy with 50x fewer parameters and< 0.5 MB model size , author=. arXiv preprint arXiv:1602.07360 , year=

    arXiv

  44. [52]

    arXiv preprint arXiv:1409.1556 , year=

    Very deep convolutional networks for large-scale image recognition , author=. arXiv preprint arXiv:1409.1556 , year=

    arXiv

  45. [53]

    Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    Deep residual learning for image recognition , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

  46. [54]

    Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    Aggregated residual transformations for deep neural networks , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

  47. [55]

    arXiv preprint arXiv:1605.07146 , year=

    Wide residual networks , author=. arXiv preprint arXiv:1605.07146 , year=

    arXiv

  48. [56]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    A convnet for the 2020s , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  49. [57]

    arXiv preprint arXiv:1805.09501 , year=

    Autoaugment: Learning augmentation policies from data , author=. arXiv preprint arXiv:1805.09501 , year=

    arXiv

  50. [58]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition workshops , pages=

    Randaugment: Practical automated data augmentation with a reduced search space , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition workshops , pages=

  51. [59]

    Proceedings of the IEEE/CVF international conference on computer vision , pages=

    Trivialaugment: Tuning-free yet state-of-the-art data augmentation , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=

  52. [60]

    Proceedings of the AAAI conference on artificial intelligence , volume=

    Random erasing data augmentation , author=. Proceedings of the AAAI conference on artificial intelligence , volume=

  53. [61]

    Proceedings of the IEEE/CVF international conference on computer vision , pages=

    Cutmix: Regularization strategy to train strong classifiers with localizable features , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=

  54. [62]

    Advances in neural information processing systems , volume=

    When does label smoothing help? , author=. Advances in neural information processing systems , volume=

  55. [63]

    Advances in neural information processing systems , volume=

    Pytorch: An imperative style, high-performance deep learning library , author=. Advances in neural information processing systems , volume=

  56. [64]

    IEEE transactions on pattern analysis and machine intelligence , volume=

    Faster R-CNN: Towards real-time object detection with region proposal networks , author=. IEEE transactions on pattern analysis and machine intelligence , volume=. 2016 , publisher=

    2016

  57. [65]

    Proceedings of the IEEE international conference on computer vision , pages=

    Mask r-cnn , author=. Proceedings of the IEEE international conference on computer vision , pages=

  58. [66]

    Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    Cascade r-cnn: Delving into high quality object detection , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

  59. [67]

    arXiv preprint arXiv:2305.09972 , year=

    Real-time flying object detection with YOLOv8 , author=. arXiv preprint arXiv:2305.09972 , year=

    arXiv

  60. [68]

    Proceedings of the European conference on computer vision (ECCV) , pages=

    Encoder-decoder with atrous separable convolution for semantic image segmentation , author=. Proceedings of the European conference on computer vision (ECCV) , pages=

  61. [69]

    Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    Pyramid scene parsing network , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

  62. [70]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Masked-attention mask transformer for universal image segmentation , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  63. [71]

    Proceedings of the IEEE/CVF international conference on computer vision , pages=

    Asymmetric non-local neural networks for semantic segmentation , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=

  64. [72]

    Proceedings of the IEEE/CVF international conference on computer vision , pages=

    Ccnet: Criss-cross attention for semantic segmentation , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=

  65. [73]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Dual attention network for scene segmentation , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  66. [74]

    Proceedings of the IEEE/CVF international conference on computer vision workshops , pages=

    Gcnet: Non-local networks meet squeeze-excitation networks and beyond , author=. Proceedings of the IEEE/CVF international conference on computer vision workshops , pages=

  67. [75]

    proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    Focal loss for dense object detection , author=. proceedings of the IEEE conference on computer vision and pattern recognition , pages=

  68. [76]

    Proceedings of the European conference on computer vision (ECCV) , pages=

    Psanet: Point-wise spatial attention network for scene parsing , author=. Proceedings of the European conference on computer vision (ECCV) , pages=

  69. [77]

    2009 IEEE conference on computer vision and pattern recognition , pages=

    Imagenet: A large-scale hierarchical image database , author=. 2009 IEEE conference on computer vision and pattern recognition , pages=. 2009 , organization=

    2009

  70. [78]

    International conference on machine learning , pages=

    Similarity of neural network representations revisited , author=. International conference on machine learning , pages=. 2019 , organization=

    2019

  71. [79]

    arXiv preprint arXiv:1906.07155 , year=

    MMDetection: Open mmlab detection toolbox and benchmark , author=. arXiv preprint arXiv:1906.07155 , year=

    1906 arXiv

  72. [80]

    MMSegmentation Contributors , howpublished =

  73. [81]

    Proceedings of the IEEE/CVF Conference on computer vision and pattern recognition , pages=

    Single-domain generalized object detection in urban scene via cyclic-disentangled self-distillation , author=. Proceedings of the IEEE/CVF Conference on computer vision and pattern recognition , pages=

  74. [82]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Object-Aware Domain Generalization for Object Detection , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  75. [83]

    arXiv preprint arXiv:2210.11466 , year=

    Surgical fine-tuning improves adaptation to distribution shifts , author=. arXiv preprint arXiv:2210.11466 , year=

    arXiv

  76. [84]

    Computer Vision--ECCV 2014: 13th European Conference, Zurich, Switzerland, September 6-12, 2014, Proceedings, Part I 13 , pages=

    Visualizing and understanding convolutional networks , author=. Computer Vision--ECCV 2014: 13th European Conference, Zurich, Switzerland, September 6-12, 2014, Proceedings, Part I 13 , pages=. 2014 , organization=

    2014

  77. [85]

    Advances in neural information processing systems , volume=

    How transferable are features in deep neural networks? , author=. Advances in neural information processing systems , volume=

  78. [86]

    Communications of the ACM , volume=

    Understanding deep learning (still) requires rethinking generalization , author=. Communications of the ACM , volume=. 2021 , publisher=

    2021

  79. [87]

    Advances in neural information processing systems , volume=

    Svcca: Singular vector canonical correlation analysis for deep learning dynamics and interpretability , author=. Advances in neural information processing systems , volume=

  80. [88]

    Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    Understanding the feature norm for out-of-distribution detection , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

  81. [89]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Improving robustness of vision transformers by reducing sensitivity to patch corruptions , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  82. [90]

    Advances in Neural Information Processing Systems , volume=

    Understanding and improving robustness of vision transformers through patch-based negative augmentation , author=. Advances in Neural Information Processing Systems , volume=

  83. [91]

    arXiv preprint arXiv:1312.6199 , year=

    Intriguing properties of neural networks , author=. arXiv preprint arXiv:1312.6199 , year=

    arXiv

  84. [92]

    Advances in Neural Information Processing Systems , volume=

    Measuring robustness to natural distribution shifts in image classification , author=. Advances in Neural Information Processing Systems , volume=

  85. [93]

    European Conference on Computer Vision , pages=

    Gabor layers enhance network robustness , author=. European Conference on Computer Vision , pages=. 2020 , organization=

    2020

  86. [94]

    International conference on machine learning , pages=

    Data determines distributional robustness in contrastive language image pre-training (clip) , author=. International conference on machine learning , pages=. 2022 , organization=

    2022

  87. [95]

    IEEE Transactions on Pattern Analysis and Machine Intelligence , year=

    Toward a holistic evaluation of robustness in clip models , author=. IEEE Transactions on Pattern Analysis and Machine Intelligence , year=

  88. [96]

    The journal of machine learning research , volume=

    Dropout: a simple way to prevent neural networks from overfitting , author=. The journal of machine learning research , volume=. 2014 , publisher=

    2014

  89. [97]

    European conference on computer vision , pages=

    Deep networks with stochastic depth , author=. European conference on computer vision , pages=. 2016 , organization=

    2016

  90. [98]

    arXiv preprint arXiv:1611.06440 , year=

    Pruning convolutional neural networks for resource efficient inference , author=. arXiv preprint arXiv:1611.06440 , year=

    arXiv

  91. [99]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Filter pruning via geometric median for deep convolutional neural networks acceleration , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  92. [100]

    The Twelfth International Conference on Learning Representations , year=

    Adaptive Sharpness-Aware Pruning for Robust Sparse Networks , author=. The Twelfth International Conference on Learning Representations , year=

  93. [101]

    International Conference on Learning Representations , year=

    Towards Deep Learning Models Resistant to Adversarial Attacks , author=. International Conference on Learning Representations , year=

  94. [102]

    IEEE transactions on pattern analysis and machine intelligence , volume=

    A survey on vision transformer , author=. IEEE transactions on pattern analysis and machine intelligence , volume=. 2022 , publisher=

    2022

  95. [103]

    Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

    Mambaout: Do we really need mamba for vision? , author=. Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

  96. [104]

    Proceedings of the IEEE/CVF international conference on computer vision , pages=

    ACDC: The adverse conditions dataset with correspondences for semantic driving scene understanding , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=

  97. [105]

    arXiv preprint arXiv:1610.01644 , year=

    Understanding intermediate layers using linear classifier probes , author=. arXiv preprint arXiv:1610.01644 , year=

    arXiv

  98. [106]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Contrastive test-time adaptation , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  99. [107]

    International Conference on Learning Representations , year=

    Tent: Fully Test-Time Adaptation by Entropy Minimization , author=. International Conference on Learning Representations , year=

  100. [108]

    Proceedings of the IEEE/CVF winter conference on applications of computer vision , pages=

    Revisiting batch normalization for improving corruption robustness , author=. Proceedings of the IEEE/CVF winter conference on applications of computer vision , pages=

  101. [109]

    International conference on machine learning , pages=

    Efficient test-time model adaptation without forgetting , author=. International conference on machine learning , pages=. 2022 , organization=

    2022

  102. [110]

    Transactions on Machine Learning Research , year=

    Santa: Source anchoring network and target alignment for continual test time adaptation , author=. Transactions on Machine Learning Research , year=

  103. [111]

    arXiv preprint arXiv:2302.12400 , year=

    Towards stable test-time adaptation in dynamic wild world , author=. arXiv preprint arXiv:2302.12400 , year=

    arXiv

  104. [112]

    The Twelfth International Conference on Learning Representations , year=

    Continual momentum filtering on parameter space for online test-time adaptation , author=. The Twelfth International Conference on Learning Representations , year=

  105. [113]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Robust mean teacher for continual and gradual test-time adaptation , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  106. [114]

    Transactions on Machine Learning Research , year=

    If your data distribution shifts, use self-learning , author=. Transactions on Machine Learning Research , year=

  107. [115]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Robust test-time adaptation in dynamic scenarios , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  108. [116]

    The Thirteenth International Conference on Learning Representations , year=

    Dynamic Sparse Training versus Dense Training: The Unexpected Winner in Image Corruption Robustness , author=. The Thirteenth International Conference on Learning Representations , year=

  109. [117]

    arXiv preprint arXiv:2010.11929 , year=

    An image is worth 16x16 words: Transformers for image recognition at scale , author=. arXiv preprint arXiv:2010.11929 , year=

    2010 arXiv

  110. [118]

    Advances in Neural Information Processing Systems , volume=

    Efficientformer: Vision transformers at mobilenet speed , author=. Advances in Neural Information Processing Systems , volume=

  111. [119]

    Proceedings of the IEEE/CVF international conference on computer vision , pages=

    Swin transformer: Hierarchical vision transformer using shifted windows , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=

  112. [120]

    arXiv preprint arXiv:2110.02178 , year=

    Mobilevit: light-weight, general-purpose, and mobile-friendly vision transformer , author=. arXiv preprint arXiv:2110.02178 , year=

    arXiv

  113. [121]

    GitHub repository , doi =

    Ross Wightman , title =. GitHub repository , doi =. 2019 , publisher =

    2019

  114. [122]

    European Conference on Computer Vision , pages=

    A lost opportunity for vision-language models: a comparative study of online test-time adaptation for vision-language models , author=. European Conference on Computer Vision , pages=. 2024 , organization=

    2024

  115. [123]

    Advances in Neural Information Processing Systems , volume=

    Intriguing properties of vision transformers , author=. Advances in Neural Information Processing Systems , volume=

Pith tools

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