Pith. sign in

REVIEW 3 major objections 6 minor 58 references

Improving Group Robustness on Spurious Correlation via Evidential Alignment

T0 review · 3 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

Pith's one-line read Evidential Alignment claims that a model's own uncertainty can locate and suppress spurious correlations without group labels, improving worst-group accuracy across vision and language benchmarks.

desk verdict A useful annotation-free debiasing method with solid empirical results, but the load-bearing uncertainty assumption is unverified and the theory is generic. read the letter →

arxiv 2506.11347 v3 pith:UM6QBFP5 submitted 2025-06-12 cs.LG

classification cs.LG
keywords spuriouscorrelationsgrouprobustnessuncertaintyquantificationevidentialdeeplearningworst-groupaccuracylast-layerretrainingsamplereweightingDirichletdistribution
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

Evidential Alignment tries to prove that the epistemic uncertainty of a deliberately biased model is enough to debias it, with no group annotations. The authors convert the usual logit classifier into a Dirichlet-evidence model via second-order risk minimization, read out per-sample uncertainty, and retrain only the last layer by upweighting the samples the biased model misclassifies in proportion to their uncertainty. On six vision and language benchmarks, they report that this reweighting lifts worst-group accuracy substantially, often matching or beating methods that use group labels for model selection. The reason to care is practical: group labels are expensive, and existing annotation-free debiasing relies on deterministic auxiliary models, while this pipeline is a lightweight extension of ordinary ERM training.

What carries the argument

The central object is the evidential Dirichlet distribution produced by second-order risk minimization. For a $K$-class problem the model outputs evidence $e_k(x)\ge 0$ and forms concentration parameters $\alpha_k(x)=e_k(x)+1$; the epistemic uncertainty readout is $u(x)=K/\sum_k\alpha_k(x)$, with high values marking samples about which the model has little evidence. The reweighting rule that carries the argument is $w(x,y)=\mathbf{1}[f_{\theta_1}(x)=y]+u(x)\mathbf{1}[f_{\theta_1}(x)\neq y]$ on a calibration set, followed by last-layer retraining with class-balanced sampling and a proximity penalty $\beta\|\theta_2-\theta_1\|_2^2$. The uncertainty readout selects which failures to amplify, and the last-layer retraining converts that selection into a debiased classifier.

What would settle it

On a split of Waterbirds or CelebA with known group labels, compute the AUROC of the stage-1 uncertainty $u(x)$ for detecting samples from the minority group among those the biased model misclassifies; if that AUROC is near 0.5, the reweighting rule is selecting noise rather than minority failures, and the method's stated mechanism is falsified even if some worst-group accuracy gain remains.

Watch

Extended reading notes

Core claim

The central discovery, on the paper's own terms, is that uncertainty estimates from a spurious-correlation-biased model carry the signal needed to identify minority-group failures: high uncertainty on misclassified calibration samples clusters in minority groups, and reweighting by $u(x)=K/\sum_k\alpha_k(x)$ suppresses the spurious correlation while preserving core features. The authors support this with an ELBO argument for second-order risk minimization, a PAC-Bayes bound showing that low reweighted empirical risk controls worst-group risk, and empirical gains on Colored MNIST, Waterbirds, CelebA, CheXpert, MultiNLI, and CivilComments across ResNet-50, ViT, and BERT backbones.

Load-bearing premise

The method works only if high uncertainty on a misclassified sample reliably marks that sample as belonging to a minority group whose spurious correlation fails; if unsure mistakes are mostly ordinary noise or hard majority cases, the reweighting pushes the model the wrong way.

Editorial extensions

If this is right

  • Reported worst-group accuracy rises from 3.74% to 84.58% on Colored MNIST, from 72.6% to 92.2% on Waterbirds, from 47.2% to 84.4% on CelebA, and reaches 73.6% on CheXpert, 74.5% on MultiNLI, and 80.2% on CivilComments.
  • Because only the last layer is retrained, the added cost is a second-order last-layer pass plus a reweighted last-layer pass, the same big-O as DFR and much cheaper than JTT's full retraining.
  • The method needs no group annotations at any stage, including model selection, where the paper uses worst-class accuracy instead of worst-group accuracy.
  • The paper claims the evidential uncertainty estimates are reliable proxies for group membership, which would let the same scores support other group-aware objectives such as balanced sampling in fully annotation-free pipelines.

Reading between the lines

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

  • Editorial inference: If uncertainty-driven reweighting is the active ingredient, the same two-stage recipe should transfer to continuous or multi-valued spurious attributes, where uncertainty readouts are aggregated over calibration samples rather than binarized by misclassification.
  • Editorial inference: The last-layer-only design leaves the feature extractor's spurious directions intact, so an extension that uses the uncertainty weights to fine-tune the full network or prune spurious embedding directions could yield larger gains, at the risk of disturbing the core features the paper says are preserved.
  • Editorial inference: The PAC-Bayes bound holds for any weighting scheme, so the paper's specific claim is that evidential uncertainty is a good proxy for the worst-group weight; comparing evidential weights with alternatives such as forgetting counts or gradient-norm outliers under the same bound would isolate that proxy's value.
  • Editorial inference: Because the method relies on the biased model's own errors, it likely inherits that model's failure modes; on datasets where majority-group samples are themselves hard or label noise is high, the uncertainty-reweighted retraining may need a denoising step.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 6 minor

Summary. The paper proposes Evidential Alignment, a two-stage annotation-free method for improving worst-group accuracy under spurious correlations. Stage 1 trains (last-layer) models with a second-order risk objective that adds a KL regularizer to a Dirichlet-evidential classification loss, yielding per-sample uncertainty estimates. Stage 2 uses a calibration set to reweight the loss for samples misclassified by the stage-1 model, weighting them by their evidential uncertainty u(x) = K/S(x), and retrains the last layer with class-balanced sampling and a proximal regularization. The authors validate the method on Colored MNIST, Waterbirds, CelebA, CheXpert, MultiNLI, and CivilComments, reporting improvements over ERM and several annotation-free baselines, and provide ablations for the KL annealing step, the weight regularization strength, and class balancing. They also present a theoretical analysis consisting of an ELBO derivation for the second-order objective and a PAC-Bayes bound for a generic reweighted empirical risk.

Significance. If the central mechanism is validated, Evidential Alignment would be a valuable addition to the annotation-free group-robustness toolbox: it is last-layer-only, computationally light, and shows consistent worst-group accuracy gains across vision and language benchmarks, with code released. The paper also includes useful backbone experiments (ViT) and calibration-set-size sensitivity. However, the paper's main conceptual contribution is the claim that evidential uncertainty identifies minority-group samples that a spurious-correlation-biased model misclassifies, and this claim is not quantitatively established. The theoretical results are generic and do not connect the uncertainty weights to group reweighting. The empirical gains are encouraging but do not by themselves verify the mechanism, so the paper currently overstates its theoretical and mechanistic support.

major comments (3)
  1. [§4.2, Eq. (14)] The reweighting rule upweights misclassified samples by u(x) = K/S(x), but a confidently wrong prediction (large S(x), small u(x)) receives a weight less than 1, i.e., it is downweighted relative to correctly classified samples. The method therefore works only if second-order risk minimization systematically assigns high uncertainty to exactly the minority-group samples that the biased model misclassifies. This is the load-bearing assumption of the paper, and it is not established. Section 5.7 reports only qualitative GradCAM and t-SNE evidence plus an unreported 'quantitative analysis showed correlations' without any numbers, AUROC, or conditioning on misclassification. Please report a quantitative measure (e.g., AUROC or correlation) of how well u(x) separates minority from majority groups, ideally restricted to the misclassified subset, and show that the result is not driven by a trivial class-imbalance artifact.
  2. [§5.5, Figure 4(c); §4.4, Theorems 4.1 and 4.2] The ablation in Figure 4(c) varies class balancing (CB) and the regularizations (Reg) but does not isolate the role of the uncertainty multiplier u(x) in Eq. (14). A critical missing baseline is a constant-weight reweighting that upweights all misclassified samples equally (u(x) = 1 for mistakes), which would show whether the uncertainty scaling contributes anything beyond JTT-style upweighting of errors. Separately, the theory in Section 4.4 does not fill this gap: Theorem 4.1 is a standard ELBO identity for evidential learning, and Theorem 4.2 is a PAC-Bayes bound for an arbitrary group weighting {w_g} that never references u(x) or the sample-level reweighting in Eq. (14). As stated, these theorems hold for any weighting scheme and therefore do not justify the claim that uncertainty-based reweighting identifies or suppresses spurious correlations. Please either add a theoretical statement that connects u(x) to group weights or soften the claim that the method is theoretically justified.
  3. [§5.4, Tables 2–4] The empirical comparison is broad, but the main text's framing that the method 'achieves the best WGA' should be more carefully scoped. On CelebA, CnC (which uses validation group labels) achieves a higher WGA (88.8) than the proposed method (84.4), and on CivilComments, DFR† with validation group labels achieves 81.8 vs. 80.2 for the proposed method. The text acknowledges these cases, but the abstract and introduction claim without qualification that the method 'significantly improves group robustness' and 'identifies and suppresses spurious correlations.' Please clarify in the comparisons and the abstract that the claimed advantage is among fully annotation-free methods, not against methods with validation set group labels.
minor comments (6)
  1. [§5.1] The CelebA description contains a typo: 'feamale' should be 'female'.
  2. [§5.5] In the ablation description, 'improves group robustness to some extend' should read 'to some extent'.
  3. [§5.2, Eq. (19)] The notation around Eq. (19) is inconsistent: the displayed WGA is defined with 1[f_theta(x)=y], but the following sentence says '1[f_theta(x) ≠ y] is the 0-1 loss'; please correct the sentence to say that the 0-1 loss is 1[f_theta(x) ≠ y] and that WGA uses the corresponding accuracy.
  4. [Algorithm 1, §4.2] The text in Section 4.2 states that samples are drawn in a class-balanced manner during evidential calibration, but the pseudocode in Algorithm 1 does not include the class-balanced sampling step; please add it for reproducibility.
  5. [References] Reference [6] appears to be mis-attributed: the cited work 'Subjective Logic: A Formalism for Reasoning Under Uncertainty' is a book by Audun Jøsang, and the author list and URL should be checked.
  6. [Tables 2–3] Several baseline entries (e.g., ERM, LfF, BPA, GEORGE) are reported without standard deviations; please add them or state that they are taken from prior papers with single runs.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the reported worst-group accuracies are external benchmarks, and the reweighting rule is an operational heuristic rather than a claim that reduces to its own inputs.

full rationale

Evidential Alignment's stage-2 reweighting (Eq. 14) uses the biased model's own misclassification indicator and Dirichlet uncertainty u(x) to reweight calibration samples, which is a bootstrap-style training strategy analogous to JTT rather than a derivation of the target quantity from itself; the final worst-group accuracy is evaluated on held-out test groups with ground-truth spurious attributes (Sec. 5.4, Eq. 19), so the main result is not definitionally forced. The theoretical results are generic: Theorem 4.1 is a standard ELBO identity for Dirichlet variational distributions, and Theorem 4.2 bounds worst-group risk for an arbitrary group weighting {w_g} without proving that the uncertainty weights realize such a weighting, so neither theorem smuggles in the conclusion. The paper's claim that uncertainty values correlate with group labels (Sec. 5.7) is asserted without reported statistics, which is an evidence-quality concern, not circularity. Self-citations [46, 51-53] appear in background or survey contexts and are not load-bearing premises. No equation can be exhibited in which a 'prediction' is identical by construction to a fitted input, so per the hard rules the appropriate finding is no significant circularity.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The paper's contribution rests on standard evidential deep learning, a generic ELBO, and a generic PAC-Bayes bound. The key extra assumption is that evidential uncertainty from a biased model identifies minority groups; that assumption is the load-bearing premise not proven by the theorems.

free parameters (2)
  • KL annealing step eta = 10 (Waterbirds), 1 (CelebA, MultiNLI, CivilComments, CheXpert)
    Controls the ramp of the evidence regularization coefficient in Eq 12; tuned per dataset using validation performance.
  • weight regularization strength beta = 10 (all datasets)
    Controls the squared-distance penalty in the calibration objective Eq 15; tuned on validation.
assumptions (4)
  • domain assumption The Dirichlet parameterization alpha_k = evidence_k + 1 from evidential deep learning gives a valid closed-form epistemic uncertainty u = K / S.
    The method relies on the evidential deep learning framework (Sensoy et al. 2018) as a valid representation of model uncertainty; this is not derived in this paper.
  • standard math Minimizing the second-order risk in Eq 12 optimizes an evidence lower bound for a variational Dirichlet posterior.
    Theorem 4.1 reproduces the standard variational ELBO argument using Bayes rule and Dirichlet KL properties.
  • ad hoc to paper There exist latent groups G pairing class labels with spurious attributes, and per-sample evidential uncertainty approximates group reweighting.
    Eq 14 replaces unknown group weights with uncertainty-weighted per-sample weights; this assumption is load-bearing and is not proven.
  • standard math The PAC-Bayes bound applies to the chosen posterior Q, prior P, and [0,1]-bounded loss.
    Theorem 4.2 is a standard union-bound PAC-Bayes argument; it does not depend on the specific uncertainty reweighting scheme.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Improving Group Robustness on Spurious Correlation via Evidential Alignment." pith.science (2026). https://pith.science/paper/UM6QBFP5

@misc{pith2026250611347,
  author       = {Pith},
  title        = {Pith review of: Improving Group Robustness on Spurious Correlation via Evidential Alignment},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UM6QBFP5}},
  note         = {Machine review of arXiv:2506.11347}
}
read the original abstract

Deep neural networks often learn and rely on spurious correlations, i.e., superficial associations between non-causal features and the targets. For instance, an image classifier may identify camels based on the desert backgrounds. While it can yield high overall accuracy during training, it degrades generalization on more diverse scenarios where such correlations do not hold. This problem poses significant challenges for out-of-distribution robustness and trustworthiness. Existing methods typically mitigate this issue by using external group annotations or auxiliary deterministic models to learn unbiased representations. However, such information is costly to obtain, and deterministic models may fail to capture the full spectrum of biases learned by the models. To address these limitations, we propose Evidential Alignment, a novel framework that leverages uncertainty quantification to understand the behavior of the biased models without requiring group annotations. By quantifying the evidence of model prediction with second-order risk minimization and calibrating the biased models with the proposed evidential calibration technique, Evidential Alignment identifies and suppresses spurious correlations while preserving core features. We theoretically justify the effectiveness of our method as capable of learning the patterns of biased models and debiasing the model without requiring any spurious correlation annotations. Empirical results demonstrate that our method significantly improves group robustness across diverse architectures and data modalities, providing a scalable and principled solution to spurious correlations.

Figures

Figures reproduced from arXiv: 2506.11347 by the authors.

Figure 1
Figure 1. An example of cow/camel classification task. Red [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Overall framework of Evidential Alignment. (a) Extract the latent embeddings and train the classifier along with [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. t-SNE visualization of the embeddings on the test set [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Ablation study on how (a) regularization strength [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]
Figure 5
Figure 5. Figure 5: GradCAM [32] visualizations for the upweighted samples and other samples in the Waterbirds dataset. 5.7 Visualization of Reweighted Samples A crucial aspect of Evidential Alignment is determining whether the model’s epistemic uncertainty can effectively reflect the err…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

58 extracted references · 40 canonical work pages

  1. [1]

    Alexander Amini, Wilko Schwarting, Ava Soleimany, and Daniela Rus. 2020. Deep evidential regression. Advances in neural information processing systems 33 (2020), 14927–14937

  2. [2]

    Martin Arjovsky, Léon Bottou, Ishaan Gulrajani, and David Lopez-Paz. 2019. Invariant risk minimization. arXiv preprint arXiv:1907.02893 (2019)

  3. [3]

    Nicholas Baker, Hongjing Lu, Gennady Erlikhman, and Philip J Kellman. 2018. Deep convolutional networks do not classify based on global object shape. PLoS computational biology 14, 12 (2018), e1006613

  4. [4]

    Solon Barocas, Moritz Hardt, and Arvind Narayanan. 2023. Fairness and machine learning: Limitations and opportunities . MIT press

  5. [5]

    Sara Beery, Grant Van Horn, and Pietro Perona. 2018. Recognition in terra incognita. In Proceedings of the European Conference on Computer Vision . 456– 473

  6. [6]

    Hooker, and Audun Jøsang

    David Bergman, André Augusto Ciré, Willem Jan van Hoeve, John N. Hooker, and Audun Jøsang. 2016. Subjective Logic: A Formalism for Reasoning Under Uncertainty. https://api.semanticscholar.org/CorpusID:86641091

  7. [7]

    Charles Blundell, Julien Cornebise, Koray Kavukcuoglu, and Daan Wierstra. 2015. Weight uncertainty in neural network. In International conference on machine learning. PMLR, 1613–1622

  8. [8]

    Daniel Borkan, Lucas Dixon, Jeffrey Sorensen, Nithum Thain, and Lucy Vasser- man. 2019. Nuanced metrics for measuring unintended bias with real data for text classification. In Companion proceedings of the 2019 world wide web conference . 491–500

Show all 58 references
  1. [9]

    Yarin Gal and Zoubin Ghahramani. 2016. Dropout as a bayesian approximation: Representing model uncertainty in deep learning. In international conference on machine learning. PMLR, 1050–1059

  2. [10]

    Karan Goel, Albert Gu, Yixuan Li, and Christopher Ré. 2020. Model patching: Closing the subgroup performance gap with data augmentation. arXiv preprint arXiv:2008.06775 (2020)

  3. [11]

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. 2016. Deep residual learning for image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition . 770–778

  4. [12]

    Eyke Hüllermeier and Willem Waegeman. 2021. Aleatoric and epistemic uncer- tainty in machine learning: An introduction to concepts and methods. Machine learning 110, 3 (2021), 457–506

  5. [13]

    Jeremy Irvin, Pranav Rajpurkar, Michael Ko, Yifan Yu, Silviana Ciurea-Ilcus, Chris Chute, Henrik Marklund, Behzad Haghgoo, Robyn Ball, Katie Shpanskaya, et al

  6. [14]

    Alex Kendall and Yarin Gal. 2017. What uncertainties do we need in bayesian deep learning for computer vision? Advances in neural information processing systems 30 (2017)

  7. [15]

    Jacob Devlin Ming-Wei Chang Kenton and Lee Kristina Toutanova. 2019. BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding. In Proceedings of NAACL-HLT. 4171–4186

  8. [16]

    Fereshte Khani and Percy Liang. 2021. Removing spurious features can hurt accuracy and affect groups disproportionately. In Proceedings of the 2021 ACM conference on fairness, accountability, and transparency . 196–205

  9. [17]

    Nayeong Kim, Sehyun Hwang, Sungsoo Ahn, Jaesik Park, and Suha Kwak. 2022. Learning debiased classifier with biased committee. Advances in Neural Informa- tion Processing Systems 35 (2022), 18403–18415

  10. [18]

    Polina Kirichenko, Pavel Izmailov, and Andrew Gordon Wilson. 2023. Last Layer Re-Training is Sufficient for Robustness to Spurious Correlations. InInternational Conference on Learning Representations

  11. [19]

    Pang Wei Koh, Shiori Sagawa, Henrik Marklund, Sang Michael Xie, Marvin Zhang, Akshay Balsubramani, Weihua Hu, Michihiro Yasunaga, Richard Lanas Phillips, Irena Gao, et al. 2021. Wilds: A benchmark of in-the-wild distribution shifts. In International Conference on Machine Learn...

  12. [20]

    Tyler LaBonte, John C Hill, Xinchen Zhang, Vidya Muthukumar, and Abhishek Kumar. 2024. The Group Robustness is in the Details: Revisiting Finetuning under Spurious Correlations. arXiv preprint arXiv:2407.13957 (2024)

  13. [21]

    Tyler LaBonte, Vidya Muthukumar, and Abhishek Kumar. 2024. Towards last- layer retraining for group robustness with fewer annotations.Advances in Neural Information Processing Systems 36 (2024)

  14. [22]

    Balaji Lakshminarayanan, Alexander Pritzel, and Charles Blundell. 2017. Simple and scalable predictive uncertainty estimation using deep ensembles. Advances in neural information processing systems 30 (2017)

  15. [23]

    Yann LeCun, Léon Bottou, Yoshua Bengio, and Patrick Haffner. 1998. Gradient- based learning applied to document recognition. Proc. IEEE 86, 11 (1998), 2278– 2324

  16. [24]

    Gaotang Li, Jiarui Liu, and Wei Hu. 2024. Bias Amplification Enhances Minority Group Performance. Transactions on Machine Learning Research (2024)

  17. [25]

    Evan Z Liu, Behzad Haghgoo, Annie S Chen, Aditi Raghunathan, Pang Wei Koh, Shiori Sagawa, Percy Liang, and Chelsea Finn. 2021. Just train twice: Improving group robustness without training group information. InInternational Conference on Machine Learning. PMLR, 6781–6792

  18. [26]

    Ziwei Liu, Ping Luo, Xiaogang Wang, and Xiaoou Tang. 2015. Deep learning face attributes in the wild. In Proceedings of the IEEE International Conference on Computer Vision. 3730–3738

  19. [27]

    Andrey Malinin and Mark Gales. 2018. Predictive uncertainty estimation via prior networks. Advances in neural information processing systems 31 (2018)

  20. [28]

    Junhyun Nam, Hyuntak Cha, Sungsoo Ahn, Jaeho Lee, and Jinwoo Shin. 2020. Learning from failure: De-biasing classifier from biased classifier. Advances in Neural Information Processing Systems 33 (2020), 20673–20684

  21. [29]

    Junhyun Nam, Jaehyung Kim, Jaeho Lee, and Jinwoo Shin. 2022. Spread Spurious Attribute: Improving Worst-group Accuracy with Spurious Attribute Estimation. In International Conference on Learning Representations

  22. [30]

    Shikai Qiu, Andres Potapczynski, Pavel Izmailov, and Andrew Gordon Wilson

  23. [31]

    Shiori Sagawa, Pang Wei Koh, Tatsunori B Hashimoto, and Percy Liang. 2020. Distributionally Robust Neural Networks. InInternational Conference on Learning Representations

  24. [32]

    Ramprasaath R Selvaraju, Michael Cogswell, Abhishek Das, Ramakrishna Vedan- tam, Devi Parikh, and Dhruv Batra. 2020. Grad-CAM: visual explanations from deep networks via gradient-based localization. International journal of computer vision 128 (2020), 336–359

  25. [33]

    Murat Sensoy, Lance Kaplan, and Melih Kandemir. 2018. Evidential deep learning to quantify classification uncertainty. Advances in neural information processing systems 31 (2018)

  26. [34]

    Seonguk Seo, Joon-Young Lee, and Bohyung Han. 2022. Unsupervised learning of debiased representations with pseudo-attributes. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . 16742–16751

  27. [35]

    Nimit Sohoni, Jared Dunnmon, Geoffrey Angus, Albert Gu, and Christopher Ré. 2020. No subclass left behind: Fine-grained robustness in coarse-grained classification problems. Advances in Neural Information Processing Systems 33 (2020), 19339–19352

  28. [36]

    Pierre Stock and Moustapha Cisse. 2018. Convnets and imagenet beyond accuracy: Understanding mistakes and uncovering biases. In Proceedings of the European Conference on Computer Vision (ECCV) . 498–512

  29. [37]

    Dennis Ulmer, Christian Hardmeier, and Jes Frellsen. 2021. Prior and poste- rior networks: A survey on evidential deep learning methods for uncertainty estimation. arXiv preprint arXiv:2110.03051 (2021)

  30. [38]

    Laurens Van der Maaten and Geoffrey Hinton. 2008. Visualizing data using t-SNE. Journal of machine learning research 9, 11 (2008)

  31. [39]

    Vladimir N Vapnik. 1999. An overview of statistical learning theory. IEEE transactions on neural networks (1999)

  32. [40]

    Jiaheng Wei, Harikrishna Narasimhan, Ehsan Amid, Wen-Sheng Chu, Yang Liu, and Abhishek Kumar. 2023. Distributionally robust post-hoc classifiers under prior shifts. arXiv preprint arXiv:2309.08825 (2023)

  33. [41]

    Welinder, S

    P. Welinder, S. Branson, T. Mita, C. Wah, F. Schroff, S. Belongie, and P. Perona

  34. [42]

    Adina Williams, Nikita Nangia, and Samuel R Bowman. 2017. A broad-coverage challenge corpus for sentence understanding through inference. arXiv preprint arXiv:1704.05426 (2017)

  35. [43]

    Yager and Liping Liu

    Ronald R. Yager and Liping Liu. 2010. Classic Works of the Dempster-Shafer Theory of Belief Functions. In Classic Works of the Dempster-Shafer Theory of Belief Functions. https://api.semanticscholar.org/CorpusID:54105362

  36. [44]

    Yuzhe Yang, Haoran Zhang, Dina Katabi, and Marzyeh Ghassemi. 2023. Change is hard: a closer look at subpopulation shift. In Proceedings of the 40th International Conference on Machine Learning . 39584–39622

  37. [45]

    Huaxiu Yao, Yu Wang, Sai Li, Linjun Zhang, Weixin Liang, James Zou, and Chelsea Finn. 2022. Improving out-of-distribution robustness via selective augmentation. In International Conference on Machine Learning . PMLR, 25407–25437

  38. [46]

    Wenqian Ye, Guangtao Zheng, Xu Cao, Yunsheng Ma, and Aidong Zhang. 2024. Spurious Correlations in Machine Learning: A Survey. http://arxiv.org/abs/2402. 12715 arXiv:2402.12715 [cs]

  39. [47]

    John R Zech, Marcus A Badgeley, Manway Liu, Anthony B Costa, Joseph J Titano, and Eric Karl Oermann. 2018. Variable generalization performance of a deep learning model to detect pneumonia in chest radiographs: a cross-sectional study. PLoS medicine 15, 11 (2018), e1002683

  40. [48]

    Jianyu Zhang, David Lopez-Paz, and Léon Bottou. 2022. Rich feature construc- tion for the optimization-generalization dilemma. In International Conference on Machine Learning. PMLR, 26397–26411

  41. [49]

    Sohoni, Hongyang R

    Michael Zhang, Nimit S. Sohoni, Hongyang R. Zhang, Chelsea Finn, and Christo- pher Re. 2022. Correct-N-Contrast: a Contrastive Approach for Improving Robust- ness to Spurious Correlations. In Proceedings of the 39th International Conference on Machine Learning. PMLR, 26484–265...

  42. [50]

    Guangtao Zheng, Wenqian Ye, and Aidong Zhang. 2024. Benchmarking Spurious Bias in Few-Shot Image Classifiers. In European Conference on Computer Vision . Springer, 346–364. Improving Group Robustness on Spurious Correlation via Evidential Alignment KDD ’25, August 3–7, 2025, T...

  43. [51]

    Guangtao Zheng, Wenqian Ye, and Aidong Zhang. 2024. Learning Robust Clas- sifiers with Self-Guided Spurious Correlation Mitigation. http://arxiv.org/abs/ 2405.03649 arXiv:2405.03649 [cs]

  44. [52]

    Guangtao Zheng, Wenqian Ye, and Aidong Zhang. 2024. Spuriousness-Aware Meta-Learning for Learning Robust Classifiers. http://arxiv.org/abs/2406.10742 arXiv:2406.10742 [cs]

  45. [53]

    Guangtao Zheng, Wenqian Ye, and Aidong Zhang. 2025. ShortcutProbe: Probing Prediction Shortcuts for Learning Robust Models. arXiv preprint arXiv:2505.13910 (2025)

  46. [54]

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

  47. [2010]

    Technical Report CNS-TR-2010-001

    Caltech-UCSD Birds 200. Technical Report CNS-TR-2010-001. California Institute of Technology

  48. [2017]

    Appendix A Dataset Distributions By examining the distributions of the four datasets, we can observe distinct patterns of group/class imbalance shown in Table 6

    Places: A 10 million image database for scene recognition.IEEE Transactions on Pattern Analysis and Machine Intelligence 40, 6 (2017), 1452–1464. Appendix A Dataset Distributions By examining the distributions of the four datasets, we can observe distinct patterns of group/cla...

  49. [2019]

    In Proceedings of the AAAI conference on artificial intelligence , Vol

    Chexpert: A large chest radiograph dataset with uncertainty labels and expert comparison. In Proceedings of the AAAI conference on artificial intelligence , Vol. 33. 590–597

  50. [2023]

    In International Conference on Machine Learning

    Simple and fast group robustness by automatic feature reweighting. In International Conference on Machine Learning . PMLR, 28448–28467

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.