Pith. sign in

REVIEW 3 major objections 4 minor 162 references

Self-supervised masking steers neural networks away from spurious features

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

AIM uses multi-stage feature guidance for self-supervised masking to improve both interpretability (EPG) and accuracy on vision benchmarks.

T0 review reviewed 2026-08-05 challenge →

load-bearing objection Clean idea with a plausible mechanism, but the abstract undersells the evidence and EPG is a shaky faithfulness proxy; worth a referee but not a citation yet. the 3 major comments →

arxiv 2508.11502 v1 pith:EUAUCUEX submitted 2025-08-15 cs.CV

AIM: Amending Inherent Interpretability via Self-Supervised Masking

Eyad Alshami , Shashank Agnihotri , Bernt Schiele , Margret Keuper This is my paper
classification cs.CV
keywords interpretabilityself-supervised learningfeature maskingspurious featuresenergy pointing gameout-of-distribution generalizationfine-grained classification
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

AIM (Amending Inherent Interpretability via Self-Supervised Masking) is a training method that makes deep neural networks rely on genuine, meaningful features instead of spurious correlations, without requiring any extra annotations. It uses features from multiple encoding stages to guide a sample-specific masking process during training. The paper reports that this simple approach consistently improves both interpretability, as measured by the Energy Pointing Game (EPG) score, and classification accuracy across diverse benchmarks including ImageNet100, HardImageNet, ImageWoof, Waterbirds, TravelingBirds, and CUB-200. If correct, AIM offers a practical way to build models that are both accurate and inherently interpretable, with better out-of-distribution generalization.

Core claim

The central claim is that a self-supervised, sample-specific feature-masking process, guided by features at multiple encoding stages, can amend a network's inherent interpretability by suppressing spurious features while preserving genuine ones. The paper demonstrates that models trained with AIM achieve significantly higher Energy Pointing Game (EPG) scores and improved accuracy compared to strong baselines across general-purpose and fine-grained classification datasets. This dual benefit holds across diverse domains and architectures, supporting the conclusion that AIM promotes the use of genuine, human-aligned features that directly contribute to better generalization and interpretability

What carries the argument

The central mechanism is a self-supervised, sample-specific masking process that uses features from multiple encoding stages of the network to decide which input regions to mask during training. This forces the model to rely less on potentially spurious cues and more on genuine, discriminative features, without needing any label-based annotation of which features are spurious.

Load-bearing premise

The method relies on the assumption that the self-supervised, multi-stage feature guidance can reliably identify and suppress spurious features while preserving genuine ones, without any external annotations or supervision about which features are spurious.

What would settle it

If, on a dataset with a known spurious correlation (e.g., Waterbirds background), running AIM with random masking instead of feature-guided masking produces the same EPG and accuracy gains, then the specific multi-stage guidance is not the cause; or, if AIM fails to improve EPG on a held-out set of images specifically chosen to expose spurious-feature reliance, the core claim would be refuted.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • Models trained with AIM are expected to produce saliency maps that more faithfully reflect the true decision process, as quantified by higher EPG scores.
  • The accuracy gains reported across datasets suggest that reducing reliance on spurious features also improves out-of-distribution generalization.
  • Because AIM requires no additional annotations, it can be applied to a wide range of existing architectures and datasets without extra labeling cost.
  • The consistent gains across general-purpose and fine-grained benchmarks indicate that the method addresses a general weakness of deep networks, not a niche artifact.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • AIM's masking may act as a form of implicit regularization that prevents shortcut learning; testing it on additional spurious-correlation benchmarks could clarify this role.
  • The multi-stage feature guidance could be adapted to other self-supervised objectives, such as contrastive learning, to inject interpretability earlier in representation learning.
  • A direct ablation replacing the feature-guided masking with random masking would test whether the specific guidance is what drives the gains, or whether any masking suffices.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 4 minor

Summary. The paper proposes AIM (Amending Inherent Interpretability via Self-Supervised Masking), a method that uses features at multiple encoding stages to create sample-specific, annotation-free masks that suppress spurious features and promote genuine ones. The abstract claims that AIM trains models that are both well-performing and inherently interpretable, yielding significant gains in Energy Pointing Game (EPG) score and accuracy across several challenging datasets (ImageNet100, HardImageNet, ImageWoof, Waterbirds, TravelingBirds, CUB-200). The submitted text contains only the abstract; no method details, equations, experimental tables, or code are available.

Significance. If the claimed results hold, AIM would be a meaningful advance toward interpretable deep learning without additional annotations, addressing both OOD generalization and human-aligned feature use. The high-level idea is plausible and the evaluation plan is broad. However, the provided material is only an abstract, and the key evidence cited (EPG) is a proxy that may not establish genuine feature use. The absence of quantitative comparisons, error bars, ablations, and an analysis of which features were masked makes it impossible to verify the central claim. The potential is real, but the current evidence is insufficient.

major comments (3)
  1. [Abstract] The claim that AIM yields models that 'faithfully summarize the decision process' is supported in the abstract only by the Energy Pointing Game (EPG) score. EPG is a pointing-game metric that measures spatial overlap between the peak of an attribution map and a human-annotated mask; a high EPG can be achieved by a model that attends to any discriminative region within the object, including a spurious cue (e.g., texture or watermark-like artifact) that lies inside the annotated area. Therefore, an EPG improvement does not, by itself, demonstrate that the model now uses 'genuine' features or that the self-supervised masking correctly identified spurious ones. A causal faithfulness metric (e.g., intervention-based tests) or a direct analysis of masked features is needed to support the central claim.
  2. [Abstract] The masking process is guided by 'features at multiple encoding stages.' If the attribution method used to compute EPG also relies on these multi-stage features, then the observed EPG gain may be an artifact of the masking sharpening the representations used by the explainer, rather than reflecting a substantive change in the model's decision behavior. The abstract does not specify what attribution method is used or whether it is independent of the masking pathway. This should be clarified, and an experiment decoupling the explainer from the masking mechanism should be reported.
  3. [Abstract] The abstract reports 'significant gains' and 'consistent gains across domains and architectures' but gives no quantitative values, no baseline names, no error bars, and no statistical tests. Without these, the magnitude and robustness of the improvements cannot be assessed. In particular, the accuracy gains on datasets such as Waterbirds are often attributed to reduced spurious correlation; the abstract does not state how much of the gain comes from the interpretability mechanism versus standard augmentation or regularization effects. Please provide numerical results and an ablation of the masking component.
minor comments (4)
  1. [Abstract] The term 'inherent interpretability' is used but not defined; it would help to distinguish it from post-hoc explainability approaches and from architectures that are interpretable by design.
  2. [Abstract] ImageWoof is a subset of ImageNet; referring to it as a general-purpose benchmark may be imprecise. Also, no references are given for the datasets or for EPG.
  3. [Abstract] The phrase 'sample-specific feature-masking process' lacks detail: what is the mask shape, are masks binary or soft, and how are they generated from multi-stage features? A sentence or two in the abstract would help the reader.
  4. [Abstract] The statement 'across domains and architectures' is vague. Which architectures were tested? One or two lines would strengthen the claim.

Circularity Check

0 steps flagged

No circularity identified from the available abstract; EPG evidence is an evidential-strength concern, not a definitional reduction.

full rationale

The available text is only the abstract; no equations, training objectives, or evaluation protocols are specified. AIM is described as using multi-stage features to guide self-supervised, sample-specific masking, with improvements measured by Energy Pointing Game (EPG) and accuracy. There is no statement that EPG is optimized during training, no fitted parameter renamed as a prediction, and no self-citation chain invoked to force a conclusion. The claim that EPG gains demonstrate faithful summarization could be questioned as an evaluation-validity issue (EPG rewards peak-overlap with human masks and may improve without causal feature use), but that is a correctness risk, not circularity. Without access to the method details, one cannot exhibit a specific reduction such as Eq. X being defined in terms of Eq. Y or a metric being optimized and then reported as a prediction. Therefore, the honest finding is no significant circularity, score 0.

Axiom & Free-Parameter Ledger

0 free parameters · 3 axioms · 0 invented entities

No additional free parameters, new entities, or untested constructs are introduced in the abstract. The assumptions listed are domain-specific premises that the method's success depends on.

axioms (3)
  • domain assumption Multi-stage features provide a reliable signal for distinguishing genuine from spurious features in a sample-specific manner.
    AIM's masking is guided by features from multiple encoding stages; without this, the method cannot select what to mask.
  • domain assumption Masking spurious features during training does not prevent learning of genuine discriminative features.
    The training objective assumes that the model can still learn from the remaining (unmasked) information.
  • domain assumption EPG score is a valid proxy for human-aligned interpretability and improves monotonically with genuine feature usage.
    The paper uses EPG as the interpretability metric; if EPG is not aligned with human judgments, the claimed interpretability benefit is weak.

reviewed 2026-08-05 · how reviews work

0 comments
Cite this review

Pith. "Pith review of AIM: Amending Inherent Interpretability via Self-Supervised Masking." pith.science (2026). https://pith.science/paper/EUAUCUEX

@misc{pith2026250811502,
  author       = {Pith},
  title        = {Pith review of: AIM: Amending Inherent Interpretability via Self-Supervised Masking},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/EUAUCUEX}},
  note         = {Machine review of arXiv:2508.11502}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

It has been observed that deep neural networks (DNNs) often use both genuine as well as spurious features. In this work, we propose "Amending Inherent Interpretability via Self-Supervised Masking" (AIM), a simple yet interestingly effective method that promotes the network's utilization of genuine features over spurious alternatives without requiring additional annotations. In particular, AIM uses features at multiple encoding stages to guide a self-supervised, sample-specific feature-masking process. As a result, AIM enables the training of well-performing and inherently interpretable models that faithfully summarize the decision process. We validate AIM across a diverse range of challenging datasets that test both out-of-distribution generalization and fine-grained visual understanding. These include general-purpose classification benchmarks such as ImageNet100, HardImageNet, and ImageWoof, as well as fine-grained classification datasets such as Waterbirds, TravelingBirds, and CUB-200. AIM demonstrates significant dual benefits: interpretability improvements, as measured by the Energy Pointing Game (EPG) score, and accuracy gains over strong baselines. These consistent gains across domains and architectures provide compelling evidence that AIM promotes the use of genuine and meaningful features that directly contribute to improved generalization and human-aligned interpretability.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

162 extracted references · 42 canonical work pages · 1 internal anchor

  1. [1]

    ICML Workshop on Uncertainty and Robustness in Deep Learning (UDL) , year=

    Towards improving robustness of compressed CNNs , author=. ICML Workshop on Uncertainty and Robustness in Deep Learning (UDL) , year=

  2. [2]

    2024 , url=

    Roll the dice: Monte Carlo Downsampling as a low-cost Adversarial Defence , author=. 2024 , url=

    2024

  3. [3]

    Synthetic Data for Computer Vision Workshop@ CVPR 2025 , year=

    Corner Cases: How Size and Position of Objects Challenge ImageNet-Trained Models , author=. Synthetic Data for Computer Vision Workshop@ CVPR 2025 , year=

    2025

  4. [4]

    arXiv preprint arXiv:2503.22399 , year=

    VITAL: More Understandable Feature Visualization through Distribution Alignment and Relevant Information Flow , author=. arXiv preprint arXiv:2503.22399 , year=

    arXiv

  5. [5]

    International conference on machine learning , pages=

    Learning transferable visual models from natural language supervision , author=. International conference on machine learning , pages=. 2021 , organization=

    2021

  6. [6]

    arXiv preprint arXiv:1606.08415 , year=

    Gaussian error linear units (gelus) , author=. arXiv preprint arXiv:1606.08415 , year=

    Pith/arXiv arXiv

  7. [7]

    2017 , url=

    Ilya Loshchilov and Frank Hutter , booktitle=. 2017 , url=

    2017

  8. [8]

    International Conference on Learning Representations , year=

    Decoupled Weight Decay Regularization , author=. International Conference on Learning Representations , year=

  9. [9]

    arXiv preprint arXiv:1412.6980 , year=

    Adam: A method for stochastic optimization , author=. arXiv preprint arXiv:1412.6980 , year=

    Pith/arXiv arXiv

  10. [10]

    What and where: A Bayesian inference theory of attention , journal =

    Sharat Chikkerur and Thomas Serre and Cheston Tan and Tomaso Poggio , keywords =. What and where: A Bayesian inference theory of attention , journal =. 2010 , note =. doi:https://doi.org/10.1016/j.visres.2010.05.013 , url =

    work page doi:10.1016/j.visres.2010.05.013 2010

  11. [11]

    Neuroreport , volume=

    Bayesian inference and attentional modulation in the visual cortex , author=. Neuroreport , volume=. 2005 , publisher=

    2005

  12. [12]

    Top-down influence in early visual processing: a Bayesian perspective , journal =

    Tai Sing Lee , keywords =. Top-down influence in early visual processing: a Bayesian perspective , journal =. 2002 , issn =. doi:https://doi.org/10.1016/S0031-9384(02)00903-4 , url =

    work page doi:10.1016/s0031-9384(02)00903-4 2002

  13. [13]

    Interactions of top-down and bottom-up mechanisms in human visual cortex

    McMains, Stephanie and Kastner, Sabine. Interactions of top-down and bottom-up mechanisms in human visual cortex. J Neurosci

  14. [14]

    Nature reviews neuroscience , volume=

    Control of goal-directed and stimulus-driven attention in the brain , author=. Nature reviews neuroscience , volume=. 2002 , publisher=

    2002

  15. [15]

    Nature reviews neuroscience , volume=

    Computational modelling of visual attention , author=. Nature reviews neuroscience , volume=. 2001 , publisher=

    2001

  16. [16]

    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=

  17. [17]

    nature , volume=

    Deep learning , author=. nature , volume=. 2015 , publisher=

    2015

  18. [18]

    International Conference on Machine Learning , pages=

    Examining and combating spurious features under distribution shift , author=. International Conference on Machine Learning , pages=. 2021 , organization=

    2021

  19. [19]

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

    Recognition in terra incognita , author=. Proceedings of the European conference on computer vision (ECCV) , pages=

  20. [20]

    International Conference on Learning Representations , year=

    Salient ImageNet: How to discover spurious features in Deep Learning? , author=. International Conference on Learning Representations , year=

  21. [21]

    arXiv preprint arXiv:2006.09994 , year=

    Noise or signal: The role of image backgrounds in object recognition , author=. arXiv preprint arXiv:2006.09994 , year=

    Pith/arXiv arXiv 2006

  22. [22]

    Nature Machine Intelligence , volume=

    Shortcut learning in deep neural networks , author=. Nature Machine Intelligence , volume=. 2020 , publisher=

    2020

  23. [23]

    arXiv preprint arXiv:1811.12231 , year=

    ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness , author=. arXiv preprint arXiv:1811.12231 , year=

    Pith/arXiv arXiv

  24. [24]

    , booktitle=

    Torralba, Antonio and Efros, Alexei A. , booktitle=. Unbiased look at dataset bias , year=

  25. [25]

    European conference on computer vision , pages=

    Revisiting visual question answering baselines , author=. European conference on computer vision , pages=. 2016 , organization=

    2016

  26. [27]

    arXiv preprint arXiv:1907.02893 , year=

    Invariant risk minimization , author=. arXiv preprint arXiv:1907.02893 , year=

    Pith/arXiv arXiv 1907

  27. [28]

    IEEE transactions on pattern analysis and machine intelligence , volume=

    Invariant scattering convolution networks , author=. IEEE transactions on pattern analysis and machine intelligence , volume=. 2013 , publisher=

    2013

  28. [29]

    arXiv preprint arXiv:1904.00760 , year=

    Approximating cnns with bag-of-local-features models works surprisingly well on imagenet , author=. arXiv preprint arXiv:1904.00760 , year=

    Pith/arXiv arXiv 1904

  29. [30]

    Proceedings of the 2021 ACM conference on fairness, accountability, and transparency , pages=

    Removing spurious features can hurt accuracy and affect groups disproportionately , author=. Proceedings of the 2021 ACM conference on fairness, accountability, and transparency , pages=

    2021

  30. [31]

    Nature communications , volume=

    Unmasking Clever Hans predictors and assessing what machines really learn , author=. Nature communications , volume=. 2019 , publisher=

    2019

  31. [32]

    http://pascallin

    The pascal visual object classes challenge,(voc2007) results , author=. http://pascallin. ecs. soton. ac. uk/challenges/VOC/voc2007/index. html. , year=

  32. [33]

    , author=

    The Clever Hans phenomenon: Communication with horses, whales, apes, and people. , author=. Annals of the New York Academy of Sciences , year=

  33. [34]

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

    Grad-cam: Visual explanations from deep networks via gradient-based localization , author=. Proceedings of the IEEE international conference on computer vision , pages=

  34. [35]

    International conference on machine learning , pages=

    Axiomatic attribution for deep networks , author=. International conference on machine learning , pages=. 2017 , organization=

    2017

  35. [36]

    International conference on machine learning , pages=

    Learning important features through propagating activation differences , author=. International conference on machine learning , pages=. 2017 , organization=

    2017

  36. [37]

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

    B-cos networks: Alignment is all we need for interpretability , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  37. [39]

    extended abstracts of the 2021 CHI conference on human factors in computing systems , pages=

    Human-AI interactive and continuous sensemaking: A case study of image classification using scribble attention maps , author=. extended abstracts of the 2021 CHI conference on human factors in computing systems , pages=

    2021

  38. [40]

    Proceedings of the ACM on Human-Computer Interaction , volume=

    Aligning eyes between humans and deep neural network through interactive attention alignment , author=. Proceedings of the ACM on Human-Computer Interaction , volume=. 2022 , publisher=

    2022

  39. [41]

    proceedings of the 28th ACM SIGKDD conference on knowledge discovery and data mining , pages=

    Res: A robust framework for guiding visual explanation , author=. proceedings of the 28th ACM SIGKDD conference on knowledge discovery and data mining , pages=

  40. [42]

    Proceedings of the 2019 AAAI/ACM Conference on AI, Ethics, and Society , pages=

    Explanatory interactive machine learning , author=. Proceedings of the 2019 AAAI/ACM Conference on AI, Ethics, and Society , pages=

    2019

  41. [43]

    Proceedings of the Workshop on Interactive Adaptive Learning (IAL 2019) , pages=

    Toward faithful explanatory active learning with self-explainable neural nets , author=. Proceedings of the Workshop on Interactive Adaptive Learning (IAL 2019) , pages=. 2019 , organization=

    2019

  42. [44]

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

    Score-CAM: Score-weighted visual explanations for convolutional neural networks , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition workshops , pages=

  43. [45]

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

    Ablation-cam: Visual explanations for deep convolutional network via gradient-free localization , author=. proceedings of the IEEE/CVF winter conference on applications of computer vision , pages=

  44. [46]

    IEEE Transactions on Image Processing , volume=

    Layercam: Exploring hierarchical class activation maps for localization , author=. IEEE Transactions on Image Processing , volume=. 2021 , publisher=

    2021

  45. [47]

    2018 IEEE winter conference on applications of computer vision (WACV) , pages=

    Grad-cam++: Generalized gradient-based visual explanations for deep convolutional networks , author=. 2018 IEEE winter conference on applications of computer vision (WACV) , pages=. 2018 , organization=

    2018

  46. [48]

    Advances in neural information processing systems , volume=

    Sanity checks for saliency maps , author=. Advances in neural information processing systems , volume=

  47. [49]

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

    Towards better understanding attribution methods , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  48. [50]

    Sanity Simulations for Saliency Methods

    Sanity simulations for saliency methods , author=. arXiv preprint arXiv:2105.06506 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  49. [51]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Do feature attribution methods correctly attribute features? , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  50. [52]

    International conference on learning representations , year=

    Post hoc explanations may be ineffective for detecting unknown spurious correlation , author=. International conference on learning representations , year=

  51. [53]

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

    Attention branch network: Learning of attention mechanism for visual explanation , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  52. [54]

    Advances in Neural Information Processing Systems , volume=

    Masktune: Mitigating spurious correlations by forcing to explore , author=. Advances in Neural Information Processing Systems , volume=

  53. [55]

    International conference on machine learning , pages=

    Interpretations are useful: penalizing explanations to align neural networks with prior knowledge , author=. International conference on machine learning , pages=. 2020 , organization=

    2020

  54. [56]

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

    On guiding visual attention with language specification , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  55. [57]

    arXiv preprint arXiv:2211.08285 , year=

    Identifying spurious correlations and correcting them with an explanation-based learning , author=. arXiv preprint arXiv:2211.08285 , year=

    Pith/arXiv arXiv

  56. [58]

    Computer Vision--ECCV 2020: 16th European Conference, Glasgow, UK, August 23--28, 2020, Proceedings, Part X 16 , pages=

    Learning what makes a difference from counterfactual examples and gradient supervision , author=. Computer Vision--ECCV 2020: 16th European Conference, Glasgow, UK, August 23--28, 2020, Proceedings, Part X 16 , pages=. 2020 , organization=

    2020

  57. [60]

    Nature Machine Intelligence , volume=

    Making deep neural networks right for the right scientific reasons by interacting with their explanations , author=. Nature Machine Intelligence , volume=. 2020 , publisher=

    2020

  58. [61]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Right for better reasons: Training differentiable models by constraining their influence functions , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  59. [63]

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

    Improving visual grounding by encouraging consistent gradient-based explanations , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  60. [64]

    Advances in neural information processing systems , volume=

    Harmonizing the object recognition strategies of deep neural networks with humans , author=. Advances in neural information processing systems , volume=

  61. [65]

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

    Studying How to Efficiently and Effectively Guide Models with Explanations , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

  62. [66]

    2024 , eprint=

    B-cos Alignment for Inherently Interpretable CNNs and Vision Transformers , author=. 2024 , eprint=

    2024

  63. [67]

    2011 , Institution =

    The Caltech-UCSD Birds-200-2011 Dataset , Author =. 2011 , Institution =

    2011

  64. [68]

    CUB-200-2011 Segmentations , DOI=

    Farrell, Ryan , year=. CUB-200-2011 Segmentations , DOI=

    2011

  65. [69]

    Places: A 10 Million Image Database for Scene Recognition , year=

    Zhou, Bolei and Lapedriza, Agata and Khosla, Aditya and Oliva, Aude and Torralba, Antonio , journal=. Places: A 10 Million Image Database for Scene Recognition , year=

  66. [70]

    International conference on machine learning , pages=

    Concept bottleneck models , author=. International conference on machine learning , pages=. 2020 , organization=

    2020

  67. [71]

    RCA engineer , volume=

    Pyramid methods in image processing , author=. RCA engineer , volume=

  68. [72]

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

    Feature pyramid networks for object detection , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

  69. [73]

    2022 , eprint=

    MaskTune: Mitigating Spurious Correlations by Forcing to Explore , author=. 2022 , eprint=

    2022

  70. [74]

    2023 , eprint=

    Masking Strategies for Background Bias Removal in Computer Vision Models , author=. 2023 , eprint=

    2023

  71. [75]

    International journal of computer vision , volume=

    Contextual priming for object detection , author=. International journal of computer vision , volume=. 2003 , publisher=

    2003

  72. [76]

    arXiv preprint arXiv:2206.07682 , year=

    Emergent abilities of large language models , author=. arXiv preprint arXiv:2206.07682 , year=

    Pith/arXiv arXiv

  73. [77]

    Nature Reviews Neuroscience , year=

    Itti, Laurent and Koch, Christof , title=. Nature Reviews Neuroscience , year=. doi:10.1038/35058500 , url=

    doi:10.1038/35058500

  74. [78]

    Frontiers in Artificial Intelligence , volume=

    Leveraging explanations in interactive machine learning: An overview , author=. Frontiers in Artificial Intelligence , volume=. 2023 , publisher=

    2023

  75. [79]

    2024 , eprint=

    Improving Visual Grounding by Encouraging Consistent Gradient-based Explanations , author=. 2024 , eprint=

    2024

  76. [80]

    Computer Vision--ECCV 2020: 16th European Conference, Glasgow, UK, August 23--28, 2020, Proceedings, Part XI 16 , pages=

    Contrastive multiview coding , author=. Computer Vision--ECCV 2020: 16th European Conference, Glasgow, UK, August 23--28, 2020, Proceedings, Part XI 16 , pages=. 2020 , organization=

    2020

  77. [81]

    Distill , year =

    Olah, Chris and Cammarata, Nick and Schubert, Ludwig and Goh, Gabriel and Petrov, Michael and Carter, Shan , title =. Distill , year =

  78. [82]

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

    Content-adaptive downsampling in convolutional neural networks , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  79. [83]

    Dynamic Convolutions: Exploiting Spatial Sparsity for Faster Inference , url=

    Verelst, Thomas and Tuytelaars, Tinne , year=. Dynamic Convolutions: Exploiting Spatial Sparsity for Faster Inference , url=. doi:10.1109/cvpr42600.2020.00239 , booktitle=

    arXiv 2020

  80. [84]

    arXiv preprint arXiv:1611.01144 , year=

    Categorical reparameterization with gumbel-softmax , author=. arXiv preprint arXiv:1611.01144 , year=

    Pith/arXiv arXiv

Showing first 80 references.

This paper was first reviewed by deepseek-v4-flash on August 5, 2026.