Machine Learning from Explanations

Jiashu Tao; Reza Shokri

REVIEW 3 major objections 4 minor 14 references

Machine Learning from Explanations

T0 review · 3 major / 4 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read Two-stage training with masks beats labels alone on small, skewed, spurious datasets

desk verdict Two-stage explanation training is a real variant, but without the Ismail baseline and with tuned eta hyperparameters, the headline gains are not yet established. read the letter →

arxiv 2507.04788 v1 pith:AFBNTHBG submitted 2025-07-07 cs.LG

Jiashu Tao , Reza Shokri This is my paper

classification cs.LG

keywords learningfromexplanationsexplanation-guidedtrainingtwo-stageoptimizationfeature-mapalignmentsmalldataregimespuriouscorrelationsclassimbalancesamplecomplexity

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper argues that labels alone are insufficient for training reliable models on small datasets, because a model can fit any arbitrary rule that agrees with the labels, including spurious ones. It proposes supplying a binary feature mask, or explanation, for each training label and alternating between a standard label-loss update and a feature-alignment update that pulls the model's internal feature map toward the features of the masked input. The claim is that this two-stage cycle makes models learn the intended reasons, yielding higher test accuracy, faster convergence, lower variance, and greater robustness to spurious correlations than label-only training. The authors demonstrate this on two synthetic geometric datasets and a real bird-species dataset, including settings with severe class imbalance and injected spurious cues.

What carries the argument

The mechanism is a two-stage optimization loop over a CNN decomposed into a feature extractor $f$, a linear mapping layer $m$, and a classifier $c$. Stage one updates all weights with the cross-entropy label loss. Stage two masks the input with the explanation to form $x' = x \otimes e(x)$, computes the softmax-normalized feature maps $\text{softmax}(f(x'))$ and $\text{softmax}(m(f(x)))$, and updates only $m$ to minimize their KL divergence. The mapping layer is trained to filter out irrelevant information and amplify the explanatory region, and updating only this layer avoids the conflicting-gradient race condition that the paper identifies in prior joint-loss methods.

What would settle it

Train the pipeline with deliberately corrupted explanations, such as masks shifted to cover only background or only a spurious square, on the same synthetic datasets; if test accuracy and reason-alignment remain at the levels achieved with correct masks, the claimed mechanism is not driving the gains, while a collapse to label-only levels would confirm that the explanations are doing the work.

Watch

Extended reading notes

Core claim

Standard empirical-risk-minimizing training on small datasets can fit any rule consistent with labels, including spurious ones. The paper's central claim is that supplying a binary mask over input features that explains each label, and alternating between a standard label-loss update and a KL-divergence update that aligns the feature map of the full input (passed through a learned mapping layer) with the feature map of the masked input, yields models that learn the intended reasons. As a result, on small, class-imbalanced, and spuriously correlated datasets, this two-stage pipeline reaches higher test accuracy faster, with lower variance and higher pairwise agreement between models trained on the same data, than label-only training, focal loss, or gradient-regularization baselines.

Load-bearing premise

The explanation masks must be informative enough to fully explain the label, and the masked input's feature map must be distinctive enough for the classifier; if a mask is incomplete, misaligned, or too coarse, the alignment step can steer the model toward partial or wrong reasons.

Editorial extensions

If this is right

Models trained with explanation masks reach a given accuracy with substantially fewer training examples, with the reported cost analysis indicating at least a 3x and sometimes a 10x reduction in sample size.
When the training set is severely class-imbalanced (9:1), the approach outperforms label-only training and focal loss, even when explanations are provided only for the minority class.
Models trained with explanations show much higher pairwise prediction agreement on the same training data, indicating that they converge to the same decision function rather than arbitrary label-consistent rules.
When spurious features are injected into training data, explanation-guided models maintain higher accuracy on clean test sets, indicating they rely more on the intended reasons.
The paper's comparison suggests that prior gradient-regularization approaches that penalize wrong reasons rarely outperform label-only training, whereas the proposed two-stage feature-alignment approach does.

Reading between the lines

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

The approach would likely transfer to knowledge distillation, where a teacher provides both labels and explanation masks to train a student model; the paper mentions this as a natural extension.
Because the second stage only updates the mapping layer, the method could be combined with pretrained feature extractors or used as a lightweight fine-tuning procedure, though the paper does not test this directly.
The alternative test sets used to probe whether models learn the intended reasons are constructed from the same explanation-derived rules, so a stronger evaluation would use independently collected human annotations or counterfactual variations not derived from the masks.
If real-world explanation masks are noisy or coarse, the KL alignment may need confidence weighting or a robust loss; the paper assumes masks are informative and sufficient, which may not hold in practice.

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.

Desk Editor's Note

read the letter

Two things to know. First, the training scheme—alternating a label step with a feature-alignment step on masked inputs, through a learned linear mapping layer—is genuinely different from the joint-loss regularization in Ross et al., Rieger et al., and Ismail et al. Second, the experiments are better than the typical small-data interpretability paper: 30 seeds, alternative test sets that reward the intended rule, spurious-feature injection, and ablations of the two new components. That is real work and it mostly supports the claim that explanations help in small-data regimes.

The strongest evidence is the alternative-rule test sets. Instead of eyeballing saliency maps, the authors construct pentagon and triangle-vs-circle sets that share the label rule, and their models transfer better. The sparse-explanation experiment, where only minority-class points have masks, is also a nice practical touch.

Now the soft spots, in severity order. The biggest is that the paper never runs the closest prior method: saliency guided training (Ismail et al., 2021). The KL loss in Section 2.3 is explicitly "similar to" that work, and Section 2.2's argument against joint losses is used to motivate the two-stage design. Without Ismail et al. in the plots, the gains attributed to two-stage alternation could actually be due to the KL feature alignment itself or to the mapping layer. The ablation in Table 3 only varies the authors' own components; it cannot separate the two-stage idea from the specific loss. Second, the hyperparameters eta_1 and eta_2 are selected from empirical results (Appendix B) with no held-out protocol. The authors even say eta_2 needed to be raised for imbalanced/spurious settings "because models need to update their focus more"—that is post hoc fitting, and it weakens the claim that the method is robust. Third, no code or data is released, and the architecture details are compressed; reproduction would be harder than it needs to be. Fourth, the table formatting in the appendix is a mess (repeated numbers, misaligned columns), which is minor but doesn't inspire confidence.

None of these are fatal. The central idea is plausible and the evidence, while incomplete, is not circular in a damaging way—the alternative test sets are constructed from the same rules as the masks, which is a mild self-consistency issue, not a logical circle. The paper should go to peer review, but only if the authors add the Ismail baseline, fix the hyperparameter story, and release code. As it stands, I would not cite it as an established result yet.

Referee Report

3 major / 4 minor

Summary. The paper proposes a two-stage training algorithm for image classifiers that uses simple explanation masks as an additional supervision signal. In stage one, all parameters are updated with a cross-entropy label loss; in stage two, only a newly inserted linear mapping layer is updated to minimize the KL divergence between the softmax-normalized feature map of the full input and the feature map of the explanation-masked input. The authors claim that this alternating schedule, unlike joint explanation-loss training, avoids gradient conflicts and leads to faster convergence, higher test accuracy, better consistency, robustness to spurious features, and reduced sample complexity in small, imbalanced, or spurious-data regimes. Experiments on two synthetic geometric datasets and a two-class CUB bird subset compare against vanilla training, gradient regularization, and focal loss, and include ablations of the mapping layer and the two-stage schedule.

Significance. If the claims hold, the paper offers a practical and low-cost way to inject human reasoning into deep learning under data scarcity. The evaluation is more thorough than typical for this line of work: it reports 30-trial statistics, includes a sparse-explanation setting, tests on alternative rule-sharing datasets, and stress-tests with injected spurious features. The two-stage optimization with a dedicated mapping layer is a plausible mechanism for avoiding the gradient conflicts that afflict joint explanation losses. However, the significance is limited by the narrow evaluation (the Bird dataset is the only real data, with hand-drawn beak masks) and by concerns about hyperparameter selection and the absence of the closest prior baseline. These issues do not invalidate the core idea but make the current evidence for the central claims incomplete.

major comments (3)

[Appendix B (hyperparameter selection)] The learning rates η1 and η2 are selected as "the best combination" from {0.1, 0.01, 0.001} based on empirical results, and the text further states that η2 is increased from 0.001 to 0.01 specifically when training under class imbalance or spurious correlations. No held-out validation set or cross-validation procedure is described for this selection, so the reported test-accuracy gains may reflect test-set tuning. Because the main claims are comparative, the authors should specify how η1 and η2 are chosen (e.g., validation split) and report the sensitivity of the results to these hyperparameters.
[Section 3 and Table 3 (baseline comparison)] The paper cites Ismail et al. (2021) as the source of the KL feature-misalignment loss but never runs their saliency guided training as a baseline. The closest evidence against joint training is the "JOINT OPT" ablation in Table 3, which uses the same loss and mapping layer but trains with a joint objective. The authors should either explicitly run and report the Ismail et al. baseline, or clearly state that the JOINT OPT condition is an instantiation of that prior method with the additional mapping layer. Without this clarification, the reader cannot attribute the observed benefits to the two-stage schedule rather than to the KL alignment objective itself.
[Section 2.3 (assumption on masked feature maps)] The method rests on the assumption that the feature map of the masked input x' = x ⊗ e(x) is "distinctive enough for the classifier," which is asserted without direct evidence. The explanation masks in the experiments are hand-crafted to highlight exactly the discriminative region (vertex, arc, or beak), so they are informative by construction. In realistic annotation settings, masks may be coarse, misaligned, or partially occluded. A brief ablation varying mask size, adding mask noise, or shifting the mask would strengthen the paper by showing the method's sensitivity to mask quality. This is load-bearing because the KL loss in Eq. (3) directly pulls the full-input feature map toward the masked-input feature map.

minor comments (4)

[Section 3.3 (alternative test sets)] The pentagon and triangle-vs-circle test sets are constructed from the same cues as the explanation masks used in training. Please acknowledge in the text that these tests verify adherence to the specific provided cue, rather than giving evidence that the model has learned a general, semantically meaningful "right reason" beyond that cue.
[Table 3 (readability)] The formatting of Table 3 is difficult to parse: many cells contain multiple numbers (e.g., "0.704±0.051 0.704±0.051 0.704±0.051 0.567±0.089 0.632±0.099"), and it is unclear which columns correspond to the three reported time points (20 epochs, 40/80 epochs, convergence). Please restructure the table so that each row and column is unambiguously labeled.
[Appendix B (contradictory statements)] The sentence "We empirically evaluated the model with three learning rates ... We then pick the best combination" is immediately followed by "Note that we did not tune these hyper-parameters to a great depth." This is contradictory and should be rephrased, since selecting the best among three values is still a form of tuning.
[Section D (cost analysis)] The cost analysis claims that explanations are more cost-efficient than collecting more data, but the arithmetic is not shown. Table 4 indicates roughly a 2-3x reduction in dataset size to reach a given accuracy, while explanations are stated to cost about 3x a label. Please make the cost comparison explicit and account for the variance in Table 4, since the conclusion is not obvious from the quoted numbers.

Circularity Check

1 steps flagged · score 2.0 of 10

Mild self-consistency in the 'learning the reasons' evaluation; the central empirical claims are not circular.

other [Section 3.3, Table 1; Sections A.1-A.2]
"we construct additional test datasets that share the same classification rules as the synthetic datasets: Pentagon Orientation Dataset ... Triangle vs Circle Dataset ... If models have learned the suggested reasons from the training datasets, they should be able to apply the same reasons to the additional datasets and reach high accuracy."

The 'suggested reasons' are the explanation masks that serve as the training signal: Section 2.1 defines explanations as 'informative enough that they can sufficiently explain the labels,' and Sections A.1-A.2 define the masks by the same vertex/arc rules used to generate the labels. The additional test sets (pentagon slice, triangle-vs-circle) are generated from those same rules. Thus a model that has internalized the explanation input will, by construction, score well on these test sets; the test measures consistency with the training signal rather than an independent generalization target. This does not make the main accuracy gains on held-out clean test sets circular, but it does make the 'learned the suggested reasons' claim partly self-referential.

full rationale

No fitted parameter is renamed as a prediction: the reported hyperparameter choices in Appendix B are ordinary tuning of learning rates, and the same tuning is applied to baselines. No load-bearing self-citation appears: the method cites external work (Ross et al., Ismail et al.) for loss components but does not rely on a self-referential uniqueness theorem or a prior paper by the same authors. The two-stage loss (Eq. 3-4, Algorithm 1) is a stated objective, not a derived result, and the central empirical claims (higher test accuracy, faster convergence, lower variance) are evaluated on held-out clean test sets that do not depend on the explanation-construction rule. The only circular element is the auxiliary evaluation in Section 3.3, where the 'alternative datasets sharing the same rules' use exactly the rules encoded in the explanation masks; that is a mild self-consistency check, not a derivation. Overall, the paper's main contribution is empirical and self-contained; the circularity is minor and localized.

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

The central claim rests on the sufficiency and correctness of explanation masks, a plausible but unproven domain assumption, and on the design choice that KL alignment of feature maps is a valid attention proxy. The method introduces no new physical entities or fitted constants beyond two tuned learning rates. Overall the approach is empirical and does not derive its target result from a circular fitting procedure.

free parameters (2)

eta_1 (learning rate for label loss) = 0.001 (synthetic and Bird)
Tuned per dataset by selecting from {0.1, 0.01, 0.001} based on empirical test performance (Section B.1, B.2); the paper states 'we pick the best combination'.
eta_2 (learning rate for feature-misalignment loss) = 0.001 (balanced, no spurious) or 0.01 (imbalanced or spurious)
Chosen from {0.1, 0.01, 0.001} using test performance; the paper reports different values for different scenarios, which is a post-hoc design choice.

assumptions (4)

domain assumption Explanations e(x) are informative and sufficient to explain the labels
Stated in Section 2.1; the entire method and its evaluation rely on this, since masked inputs are used as the target for the feature-misalignment loss.
domain assumption Latent representations of the masked inputs are distinctive enough for classification
Stated in Section 2.3; if false, aligning feature maps to them will not yield a useful classifier.
ad hoc to paper KL divergence between softmax-normalized feature maps is a faithful proxy for attention alignment
Design choice in Equation (3); no theoretical or empirical justification is given that reducing this KL forces the model to rely on the highlighted features in the intended causal way.
ad hoc to paper Updating only the mapping layer in the second stage avoids gradient conflicts
Claimed in Section 2.3 and ablation C, but the reasoning is intuitive; there is no analysis of the loss landscape or proof that alternating updates do not introduce their own instability.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Machine Learning from Explanations." pith.science (2026). https://pith.science/paper/AFBNTHBG

@misc{pith2026250704788,
  author       = {Pith},
  title        = {Pith review of: Machine Learning from Explanations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/AFBNTHBG}},
  note         = {Machine review of arXiv:2507.04788}
}

read the original abstract

Acquiring and training on large-scale labeled data can be impractical due to cost constraints. Additionally, the use of small training datasets can result in considerable variability in model outcomes, overfitting, and learning of spurious correlations. A crucial shortcoming of data labels is their lack of any reasoning behind a specific label assignment, causing models to learn any arbitrary classification rule as long as it aligns data with labels. To overcome these issues, we introduce an innovative approach for training reliable classification models on smaller datasets, by using simple explanation signals such as important input features from labeled data. Our method centers around a two-stage training cycle that alternates between enhancing model prediction accuracy and refining its attention to match the explanations. This instructs models to grasp the rationale behind label assignments during their learning phase. We demonstrate that our training cycle expedites the convergence towards more accurate and reliable models, particularly for small, class-imbalanced training data, or data with spurious features.

Figures

Figures reproduced from arXiv: 2507.04788 by the authors.

**Figure 1.** A demonstration of how our proposed training pipeline uses the explanation. Solid line: First stage of the optimization, forward pass, and backpropagation of the cross entropy loss in the same way as the conventional machine learning training pipeline. In this example, an image x passes through the convolutional layers, which are the feature extractor, the self mapping layer, and the fully connected (FC) layers to r… view at source ↗

**Figure 2.** Test performance on the Triangle Orientation dataset. Top row: Training with small sample sizes with balanced classes. Bottom row: Training with small sample sizes when positive data only make up 10% of the training set. only training with explanations when the data are from the minority class (Class 1). The cyan lines in [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗

**Figure 3.** Test performance on the Fox vs Cat dataset. Top row: Training with small sample sizes with balanced classes. Bottom row: Training with small sample sizes when positive data only make up 10% of the training set. 0 100 200 0.5 0.6 0.7 0.8 0.9 Epoch TestAccuracy 30 Samples 0 100 200 0.5 0.6 0.7 0.8 0.9 Epoch 60 Samples No Expl Grad Reg Ours [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗

Figures from the paper (6 more)

**Figure 4.** Figure 4: Test performance on the Bird dataset. dataset. The vertical slice is a pentagon. If the triangle points upwards, the pentagon should also point up. • Triangle vs Circle Dataset: We construct this dataset by removing the rectangles in the test set of the Fox vs Cat data…

**Figure 5.** Figure 5: Test accuracy on the clean test set when models are trained on a spurious training set. Our models consistently achieve the highest accuracy, implying that they have learned more from the suggested reasons. suggested by the given explanations, which subsequently explai…

**Figure 6.** Figure 6: Fox (Left) vs Cat (Right). A.3. Bird We also extend our experiments to real datasets. We use a subset of the CUB-200-2011 (We will refer to it as the Bird dataset in the rest of the paper) dataset (Wah et al., 2011), a fine-grained classification dataset with photos of…

**Figure 7.** Figure 7: Indigo Buntings (Left) vs Blue Grosbeaks (Right). The difference is that Blue Grosbeaks have larger breaks. The explanations hence highlight their beaks. (a) Spurious cat (b) Spurious triangle pointing down (c) Spurious Indigo Bunting (d) Spurious fox (e) Spurious tria…

**Figure 8.** Figure 8: Top row: images with spurious features injected from the negative class. Bottom row: images with spurious features injected from the positive class. Left: a square is added to the top right corner for cat images and the top left corner for fox images. Mid: a square is …

**Figure 9.** Figure 9: Test accuracy on test sets sharing the same true reasons when models are trained on spurious training set. Our models always do better, suggesting they have learned more of the true reasons [PITH_FULL_IMAGE:figures/full_fig_p016_9.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

14 extracted references · 7 canonical work pages

[2]

Optimizing Relevance Maps of Vision Transformers Improves Robustness

Chefer, H., Schwartz, I., and Wolf, L. Optimizing relevance maps of vision transformers improves robustness.arXiv preprint arXiv:2206.01161,

work page Pith review arXiv
[5]

Learning credible deep neural networks with rationale regularization

Du, M., Liu, N., Yang, F., and Hu, X. Learning credible deep neural networks with rationale regularization. In 2019 IEEE International Conference on Data Mining (ICDM), pp. 150–159. IEEE,

work page 2019
[8]

Swin transformer v2: Scaling up capacity and resolution.arXiv preprint arXiv:2111.09883,

Liu, Z., Hu, H., Lin, Y ., Yao, Z., Xie, Z., Wei, Y ., Ning, J., Cao, Y ., Zhang, Z., Dong, L., et al. Swin transformer v2: Scaling up capacity and resolution.arXiv preprint arXiv:2111.09883,

arXiv
[10]

Do input gradients highlight discriminative features?Advances in Neural Information Processing Systems, 34:2046–2059,

Shah, H., Jain, P., and Netrapalli, P. Do input gradients highlight discriminative features?Advances in Neural Information Processing Systems, 34:2046–2059,

work page
[12]

Coping with Label Shift via Distributionally Robust Optimisation

Zhang, J., Menon, A., Veit, A., Bhojanapalli, S., Kumar, S., and Sra, S. Coping with label shift via distributionally robust optimisation.arXiv preprint arXiv:2010.12230, 2020a. Zhang, Y ., Qin, J., Park, D. S., Han, W., Chiu, C.-C., Pang, R., Le, Q. V ., and Wu, Y . Pushing the limits of semi- supervised learning for automatic speech recognition. arXiv p...

work page Pith review arXiv 2010
[13]

The explanation is then represented as a mask highlighting the vertex and a portion of the bottom line

The explanation is the spatial arrangement of the vertex and the bottom line: if the vertex is above the bottom line, it is pointing upwards. The explanation is then represented as a mask highlighting the vertex and a portion of the bottom line. Each image is of grayscale with a size of64. A.2. Fox vs Cat The second synthetic dataset is a minimalistic fox...

work page 2011
[14]

12 Machine Learning from Explanations B

Top row:images with spurious features injected from the negative class.Bottom row:images with spurious features injected from the positive class.Left:a square is added to the top right corner for cat images and the top left corner for fox images.Mid:a square is added to the top left corner for triangles pointing down and the bottom left corner for triangl...

work page 2017
[2016]

The caltech-ucsd birds-200-2011 dataset

Wah, C., Branson, S., Welinder, P., Perona, P., and Belongie, S. The caltech-ucsd birds-200-2011 dataset. Technical Report CNS-TR-2011-001, California Institute of Tech- nology,

work page 2011

Show all 14 references

[2017]

W., Hashimoto, T

Sagawa, S., Koh, P. W., Hashimoto, T. B., and Liang, P. Distributionally robust neural networks for group shifts: On the importance of regularization for worst-case gener- alization.arXiv preprint arXiv:1911.08731,

1911 arXiv
[2018]

Invariant risk minimization.arXiv preprint arXiv:1907.02893,

Arjovsky, M., Bottou, L., Gulrajani, I., and Lopez- Paz, D. Invariant risk minimization.arXiv preprint arXiv:1907.02893,

1907 arXiv
[2019]

A typology to explore and guide explanatory interac- tive machine learning.arXiv preprint arXiv:2203.03668,

Friedrich, F., Stammer, W., Schramowski, P., and Kersting, K. A typology to explore and guide explanatory interac- tive machine learning.arXiv preprint arXiv:2203.03668,

arXiv
[2020]

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

Dosovitskiy, A., Beyer, L., Kolesnikov, A., Weissenborn, D., Zhai, X., Unterthiner, T., Dehghani, M., Minderer, M., Heigold, G., Gelly, S., et al. An image is worth 16x16 words: Transformers for image recognition at scale.arXiv preprint arXiv:2010.11929,

2010 arXiv
[2021]

Kirichenko, P., Izmailov, P., and Wilson, A. G. Last layer re-training is sufficient for robustness to spurious correla- tions.arXiv preprint arXiv:2204.02937,

arXiv
[2022]

D., et al

D’Amour, A., Heller, K., Moldovan, D., Adlam, B., Ali- panahi, B., Beutel, A., Chen, C., Deaton, J., Eisenstein, J., Hoffman, M. D., et al. Underspecification presents challenges for credibility in modern machine learning. arXiv preprint arXiv:2011.03395,

2011 arXiv

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.

[1] [2]

Optimizing Relevance Maps of Vision Transformers Improves Robustness

Chefer, H., Schwartz, I., and Wolf, L. Optimizing relevance maps of vision transformers improves robustness.arXiv preprint arXiv:2206.01161,

work page Pith review arXiv

[2] [5]

Learning credible deep neural networks with rationale regularization

Du, M., Liu, N., Yang, F., and Hu, X. Learning credible deep neural networks with rationale regularization. In 2019 IEEE International Conference on Data Mining (ICDM), pp. 150–159. IEEE,

work page 2019

[3] [8]

Swin transformer v2: Scaling up capacity and resolution.arXiv preprint arXiv:2111.09883,

Liu, Z., Hu, H., Lin, Y ., Yao, Z., Xie, Z., Wei, Y ., Ning, J., Cao, Y ., Zhang, Z., Dong, L., et al. Swin transformer v2: Scaling up capacity and resolution.arXiv preprint arXiv:2111.09883,

arXiv

[4] [10]

Do input gradients highlight discriminative features?Advances in Neural Information Processing Systems, 34:2046–2059,

Shah, H., Jain, P., and Netrapalli, P. Do input gradients highlight discriminative features?Advances in Neural Information Processing Systems, 34:2046–2059,

work page

[5] [12]

Coping with Label Shift via Distributionally Robust Optimisation

Zhang, J., Menon, A., Veit, A., Bhojanapalli, S., Kumar, S., and Sra, S. Coping with label shift via distributionally robust optimisation.arXiv preprint arXiv:2010.12230, 2020a. Zhang, Y ., Qin, J., Park, D. S., Han, W., Chiu, C.-C., Pang, R., Le, Q. V ., and Wu, Y . Pushing the limits of semi- supervised learning for automatic speech recognition. arXiv p...

work page Pith review arXiv 2010

[6] [13]

The explanation is then represented as a mask highlighting the vertex and a portion of the bottom line

The explanation is the spatial arrangement of the vertex and the bottom line: if the vertex is above the bottom line, it is pointing upwards. The explanation is then represented as a mask highlighting the vertex and a portion of the bottom line. Each image is of grayscale with a size of64. A.2. Fox vs Cat The second synthetic dataset is a minimalistic fox...

work page 2011

[7] [14]

12 Machine Learning from Explanations B

Top row:images with spurious features injected from the negative class.Bottom row:images with spurious features injected from the positive class.Left:a square is added to the top right corner for cat images and the top left corner for fox images.Mid:a square is added to the top left corner for triangles pointing down and the bottom left corner for triangl...

work page 2017

[8] [2016]

The caltech-ucsd birds-200-2011 dataset

Wah, C., Branson, S., Welinder, P., Perona, P., and Belongie, S. The caltech-ucsd birds-200-2011 dataset. Technical Report CNS-TR-2011-001, California Institute of Tech- nology,

work page 2011