Pith. sign in

REVIEW 3 major objections 3 minor 30 references

Toward Simple and Robust Contrastive Explanations for Image Classification by Leveraging Instance Similarity and Concept Relevance

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

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

Pith's one-line read Contrastive explanations for image classifiers get shorter as concept relevance rises, and their length is stable under a 180-degree rotation but not under a 10-degree rotation or Gaussian noise.

desk verdict The main claim about relevance and explanation length is likely baked into the binning scheme; the robustness comparison is useful but contains a factual misreading of the noise results. read the letter →

arxiv 2506.23975 v1 pith:D5OXZ4RP submitted 2025-06-30 cs.CV

Yuliia Kaidashova , Bettina Finzel , Ute Schmid This is my paper · ORCID
classification cs.CV
keywords contrastiveexplanationconcept-basedcomplexityrobustnessinstancesimilarityconceptrelevanceimageclassificationPropagation
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

This paper proposes and evaluates a method for contrastive explanations of image classifiers that answers 'why this class rather than that one?' using human-understandable concepts. The central claim is that the concepts a model ranks as highly relevant produce shorter, less complex explanations, and that explanation length is mostly stable under a 180-degree rotation but not under a 10-degree rotation or Gaussian noise. The experiments, run on a fine-tuned VGG16 classifier distinguishing teapots from vases, show median explanation length rising from about 17 concepts in the top relevance range to about 100 in the bottom range, with an ANOVA F-value of 354.41. The robustness tests show the top relevance ranges surviving a 180-degree flip ($p > 0.05$) while Gaussian noise lengthens explanations significantly in every relevance range. The work matters because short, stable explanations are a prerequisite for trusting a model's stated reasons in real-world image classification.

What carries the argument

The central mechanism is the instance-similarity foil selection: for each explained image, Algorithm 1 chooses as the contrast the single opposite-class image whose embedding has the maximum cosine similarity, then differences the two images' concept sets. The explanation is the set of unique concepts—those appearing in only one of the two images—rendered as a natural-language sentence of the form 'classified as teapot instead of vase because it contains ... and does not contain ...'. Concept relevance comes from Concept Relevance Propagation, which isolates the neurons associated with a concept through a masked backward pass; the paper then sorts concepts by relevance and cuts them into four quartile ranges based on cumulative relevance. Explanation length, the count of concepts in the unique set, is the paper's operational measure of complexity, and it is the quantity compared across relevance ranges and across original versus augmented images.

What would settle it

Re-run the explanation algorithm on the same test images while varying the foil: use the second-most-similar image, a random image from the contrast class, or an average over several near neighbours, and recompute explanation lengths within the four relevance ranges. If the monotone increase from very strong to very low relevance disappears, reverses, or becomes non-significant under alternative foil selection, then the paper's central length result is an artifact of the single nearest-neighbor pairing rather than a general property of concept relevance.

Watch

Extended reading notes

Core claim

On its own terms, the paper discovers a monotone relation between concept relevance and explanation complexity: explanations assembled from very strong concepts are short and focused, while explanations from very low relevance concepts are long and diffuse. For each target image, the method picks the contrast-class image whose embedding is most cosine-similar, extracts per-concept relevance scores for both images with Concept Relevance Propagation, and keeps only the unique concepts that appear in one image but not the other; explanation length is the number of those concepts. Grouping concepts by cumulative relevance quartiles (top 25%, 25–50%, 50–75%, bottom 25%) yields median lengths of roughly 17, 40, 78, and 100, and the ANOVA test confirms the differences are not chance. The robustness finding is nuanced: a 180-degree rotation leaves the top three relevance ranges statistically unchanged (only the very low range shifts, $p = 0.0002$), while a 10-degree rotation shortens explanations in the strong and low ranges, and Gaussian noise significantly lengthens explanations in all four ranges. The paper interprets this as support for $H_1$ and partial support for $H_2$, and frames the noise sensitivity as the model losing focus on a small discriminative concept set.

Load-bearing premise

The load-bearing assumption is that the single opposite-class image the model judges most similar is a representative foil; if a different foil image were chosen, the set of unique concepts and every measured explanation length could change, and with them the R1 and R2 results.

Editorial extensions

If this is right

  • Restricting explanations to the top relevance quartile can reduce median explanation length from about 100 concepts to about 17 while still reporting the same contrast.
  • Explanation length can serve as a measurable, model-internal proxy for explanation complexity in concept-based methods.
  • Robustness evaluation should include small geometric perturbations: a 10-degree rotation changes explanation length in the strong and low relevance ranges, while a 180-degree rotation leaves the top ranges unchanged.
  • Gaussian noise should be treated as a serious stress test for concept-based explanation stability, since it significantly lengthens explanations in every relevance range.

Reading between the lines

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

  • An implication the paper leaves implicit is that the entire length distribution rests on the choice of a single foil image; varying the foil selection, such as averaging over several near neighbours, would show whether the monotone trend is a property of concept relevance or an artifact of that one nearest-neighbor pairing.
  • The noise result suggests explanation length could double as a detector of model uncertainty or shortcut reliance, but the paper does not test this link.
  • The teapot-versus-vase setup could be extended to other visually similar class pairs; if the quartile length ordering replicates, the method would be a general recipe for simple contrastive explanations rather than a dataset-specific observation.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 3 minor

Summary. The paper proposes a concept-based contrastive explanation method for image classification that pairs a target instance with the most cosine-similar instance from a contrast class, extracts concept relevance scores via Concept Relevance Propagation (CRP), and reports as an explanation the concepts unique to one of the two compared images. The method is evaluated along two axes: (R1) whether explanation length varies across relevance ranges defined as cumulative-relevance quantiles, and (R2) whether explanation length remains stable under Gaussian noise, 10-degree rotation, and 180-degree rotation. The authors report that higher-relevance concepts yield shorter explanations (H1 supported) and that robustness depends strongly on the augmentation type (H2 partially supported). The central positive claim is, however, undermined by the way relevance ranges are defined, and the discussion of the Gaussian noise result directly contradicts the tabulated and plotted results.

Significance. If the R1 finding were valid, the paper would make a useful empirical contribution to concept-based explainability, showing that relevance strength is a lever on explanation conciseness. The implementation is a concrete instantiation of earlier conceptual work (Finzel et al.), and the choice of a semantically close contrast pair is a reasonable design. However, the main statistical claim is potentially a definitional artifact of cumulative-relevance binning rather than an empirical property of the explanation method. The noise-robustness finding is also misdescribed in the Discussion, and the foil-selection strategy is not stress-tested. The paper's significance therefore depends on a re-analysis that separates bin-size effects from true relevance effects.

major comments (3)
  1. [Section 4] The primary result H1 is likely an artifact of the binning scheme. Relevance ranges are defined as cumulative-relevance quantiles: 'very strong' is the top 25% of total relevance, 'strong' is the next 25%, and so on. In any skewed relevance distribution, the top 25% of total relevance is concentrated in few concepts while the bottom 25% spans many concepts. Because explanation length is the number of unique concepts produced by Algorithm 1 (lines 16-18), the reported medians of approximately 17, 40, 78, and 100 closely mirror the concept counts per bin rather than any contrastive property of the explanation. The ANOVA F-value of 354.41 only shows that the four constructed groups differ in mean length; it cannot distinguish a relevance effect from a mechanical bin-size effect. The manuscript never reports the number of concepts in each relevance range or the shape of the relevance distribution. To support H1, the authors should either use fixed-size concept bins (e.g., the top 10 concepts vs. the next 10 concepts) or normalize explanation length by the number of concepts in each bin, and show that the monotonic trend survives.
  2. [Section 5] The Discussion misstates the Gaussian noise result. Section 4 states that 'across all relevance levels, explanation lengths are consistently and significantly lower for noisy images compared to the original ones,' and Figure 6 shows downward shifts in the distributions. Section 5, however, claims that 'Explanation lengths increased significantly across all relevance ranges when noise was added' and interprets this as the model losing focus on a small set of discriminative concepts. Table 1 confirms the Section 4 description with positive t-values (e.g., t=25.7762 for very strong relevance), indicating shorter explanations under noise. This internal contradiction directly affects the interpretation of H2 and the robustness conclusions; the Discussion must be corrected to describe the actual direction of the effect.
  3. [Algorithm 1] The explanation length and all downstream results depend on the selection of exactly one contrastive image, namely the embedding-nearest neighbor from the opposite class. The algorithm does not vary this selection, compare against random foils, or test sensitivity to the similarity criterion. Because uniqueness is defined against this single foil, the sets of unique concepts and hence all measured lengths in R1 and R2 could change markedly if a different foil image were used. The authors should at least report a sensitivity analysis (e.g., lengths obtained with the second- and third-nearest neighbors, or with a random foil distribution) to establish that the reported trends are not specific to one arbitrarily chosen pairing.
minor comments (3)
  1. [Table 1] The p-values are reported as '0.0000'; this should read '<0.0001' or similar to avoid implying an exact zero probability.
  2. [Section 2] There is a spacing typo in the phrase 'whereP is the observed fact' twice; 'whereP' should be 'where P'.
  3. [Section 4] The statement 'The maximum possible explanation length is 512' is not derived or justified; the maximum depends on the number of extractable concepts in the chosen CRP layer and how uniqueness is counted, so it should be either formally derived or stated as an upper bound with explanation.

Circularity Check

1 steps flagged · score 7.0 of 10

R1's 'higher relevance -> shorter explanations' is largely a definitional consequence of cumulative-relevance binning; H2 robustness results remain empirical.

  1. self definitional [Section 4 (R1 evaluation), relevance-range definition and Fig. 3]
    "To compute the ranges, all concepts are first sorted in descending order of relevance for a given prediction, where very strong relevance is the top 25% of total relevance, strong are concepts that lie between 25% and 50%, low are concepts that account for the 50% to 75% and very low is the remaining 25% of the cumulative relevance. ... For very strong relevance, the median explanation length is around 17. ... For very low relevance, we can see that the median value of explanation length is around 100."

    The relevance ranges are cumulative-relevance quantiles over concepts sorted in descending relevance. For any non-increasing relevance vector, the number of concepts needed to accumulate each successive quarter of total relevance is non-decreasing, so the 'very strong' bin contains the fewest concepts and the 'very low' bin the most by construction. The R1 outcome, explanation length, is the number of concepts in the explanation for that range (Fig. 3). Thus the reported monotone medians (about 17, 40, 78, 100) mostly mirror the mechanically increasing bin sizes.

full rationale

The circularity is confined to R1/H1. The relevance ranges are not independent treatment groups: they are constructed by taking the top 25% of cumulative relevance, then the next 25%, and so on, so the top range contains few concepts and the bottom range contains many whenever the relevance scores are non-increasing. Explanation length is measured by the number of concepts in the explanation for that range, so the monotonic trend in median lengths is largely a built-in property of the binning scheme. The paper never reports the number of concepts per relevance range or a fixed-size-bin control, and the ANOVA cannot distinguish the mechanical bin-size effect from any genuine relevance effect. This is not a self-citation-chain problem: the definitional reduction is fully visible in Section 4's own definitions, independent of the authors' earlier papers [9,10]. R2/H2, by contrast, is an empirical comparison of original versus augmented explanations using paired t-tests; those results do not reduce by construction and remain self-contained evidence. The foil-selection issue in Algorithm 1 is a representativeness limitation, not circularity, because choosing the nearest neighbor does not by itself force the measured length trend. Overall, one central prediction (H1) is largely definitional while H2 retains independent empirical content, yielding partial circularity.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

No new entities are introduced. The method reuses CRP concepts and embeddings from prior work. The free parameters are the relevance-range boundaries, the CRP layer, and augmentation parameters; the relevance-range boundaries directly shape the main finding.

free parameters (3)
  • Relevance range boundaries = 25%, 50%, 75% of cumulative relevance
    Chosen by hand to define 'very strong', 'strong', 'low', and 'very low'; the monotonic R1 trend follows from this cumulative-quantile construction.
  • CRP layer index = 40
    Selected for concept extraction; no ablation or justification is provided.
  • Augmentation parameters = 10 and 180 degrees; Gaussian noise sigma not specified
    Rotation angles and noise level are chosen by hand; noise parameters are not reported, affecting reproducibility of robustness results.
assumptions (5)
  • domain assumption Shorter explanations are preferable for human understanding.
    Invoked in Section 4 and Discussion; supported by prior work, but not evaluated with human subjects in this paper.
  • domain assumption CRP concept relevance scores faithfully reflect the model's decision process and are human-understandable.
    The whole pipeline depends on CRP (Achtibat et al.) and the mapping of neurons to concepts; no verification that layer 40 concepts are semantically meaningful.
  • domain assumption The single nearest-neighbor embedding is a representative contrastive foil.
    Algorithm 1 picks only the maximum cosine-similarity instance from the opposite class; all unique-concept sets are derived from this single pair.
  • ad hoc to paper Cumulative relevance quantiles are a valid way to group concepts into relevance levels.
    The four ranges are defined as quantiles of cumulative relevance; this construction is what makes the R1 trend appear.
  • standard math ANOVA and paired t-test assumptions (normality, independence) hold for explanation lengths.
    Statistical tests are applied to non-independent, bounded count data; no corrections or assumption checks are reported.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Toward Simple and Robust Contrastive Explanations for Image Classification by Leveraging Instance Similarity and Concept Relevance." pith.science (2026). https://pith.science/paper/D5OXZ4RP

@misc{pith2026250623975,
  author       = {Pith},
  title        = {Pith review of: Toward Simple and Robust Contrastive Explanations for Image Classification by Leveraging Instance Similarity and Concept Relevance},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/D5OXZ4RP}},
  note         = {Machine review of arXiv:2506.23975}
}
read the original abstract

Understanding why a classification model prefers one class over another for an input instance is the challenge of contrastive explanation. This work implements concept-based contrastive explanations for image classification by leveraging the similarity of instance embeddings and relevance of human-understandable concepts used by a fine-tuned deep learning model. Our approach extracts concepts with their relevance score, computes contrasts for similar instances, and evaluates the resulting contrastive explanations based on explanation complexity. Robustness is tested for different image augmentations. Two research questions are addressed: (1) whether explanation complexity varies across different relevance ranges, and (2) whether explanation complexity remains consistent under image augmentations such as rotation and noise. The results confirm that for our experiments higher concept relevance leads to shorter, less complex explanations, while lower relevance results in longer, more diffuse explanations. Additionally, explanations show varying degrees of robustness. The discussion of these findings offers insights into the potential of building more interpretable and robust AI systems.

Figures

Figures reproduced from arXiv: 2506.23975 by the authors.

Figure 1
Figure 1. Illustration of the overall concept-based contrastive explanation approach [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. Contrastive explanation for teapot and vase (best viewed in color) [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Explanation length across relevance ranges (n = 615 teapots, 328 vases) [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Explanation length for Original vs. 180-degree Rotation (n = 615 teapots, [PITH_FULL_IMAGE:figures/full_fig_p016_4.png]
Figure 5
Figure 5. Figure 5: Explanation length for Original vs. 10-degree Rotation (n = 615 teapots, [PITH_FULL_IMAGE:figures/full_fig_p017_5.png]
Figure 6
Figure 6. Figure 6: Explanation length for Original vs. Gaussian Noise (n = 615 teapots, 328 [PITH_FULL_IMAGE:figures/full_fig_p018_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

30 extracted references · 12 canonical work pages

  1. [1]

    Achtibat, R., Dreyer, M., Eisenbraun, I., Bosse, S., Wiegand, T., Samek, W., Lapuschkin, S.: From attribution maps to human-understandable explanations through concept relevance propagation. Nat. Mach. Intell.5(9), 1006–1019 (2023). https://doi.org/https://doi.org/10.1038/s42256-023-00711-8

    doi:10.1038/s42256-023-00711-8 2023

  2. [2]

    IEEE Access 6, 52138–52160 (2018)

    Adadi, A., Berrada, M.: Peeking inside the black-box: A survey on ex- plainable artificial intelligence (XAI). IEEE Access 6, 52138–52160 (2018). https://doi.org/10.1109/ACCESS.2018.2870052

    arXiv 2018

  3. [3]

    Ali, S., Abuhmed, T., El-Sappagh, S.H.A., Muhammad, K., Alonso-Moral, J.M., Confalonieri, R., Guidotti, R., Ser, J.D., Rodríguez, N.D., Herrera, F.: Explainable artificial intelligence (XAI): what we know and what is left to attain trustworthy artificial intelligence. Inf. Fusion 99, 101805 (2023). https://doi.org/10.1016/J.INFFUS.2023.101805

    arXiv 2023

  4. [4]

    In: 2020 IEEE Inter- national Conference On Artificial Intelligence Testing (AITest)

    Arcaini, P., Bombarda, A., Bonfanti, S., Gargantini, A.: Dealing with robustness of convolutional neural networks for image classification. In: 2020 IEEE Inter- national Conference On Artificial Intelligence Testing (AITest). pp. 7–14 (2020). https://doi.org/10.1109/AITEST49225.2020.00009

    work page arXiv 2020

  5. [5]

    Frontiers Artif

    Bruckert, S., Finzel, B., Schmid, U.: The next generation of medical decision sup- port: A roadmap toward transparent expert companions. Frontiers Artif. Intell.3, 507973 (2020). https://doi.org/10.3389/FRAI.2020.507973

    work page arXiv 2020

  6. [6]

    In: 2009 IEEE Computer Society Confer- ence on Computer Vision and Pattern Recognition (CVPR 2009), 20-25 June 2009, Miami, Florida, USA

    Deng, J., Dong, W., Socher, R., Li, L., Li, K., Fei-Fei, L.: Imagenet: A large- scale hierarchical image database. In: 2009 IEEE Computer Society Confer- ence on Computer Vision and Pattern Recognition (CVPR 2009), 20-25 June 2009, Miami, Florida, USA. pp. 248–255. IEEE Computer Society (2009). https://doi.org/10.1109/CVPR.2009.5206848

    arXiv 2009

  7. [7]

    In: Bengio, S., Wallach, H.M., Larochelle, H., Grauman, K., Cesa- Bianchi, N., Garnett, R

    Dhurandhar, A., Chen, P., Luss, R., Tu, C., Ting, P., Shanmugam, K., Das, P.: Explanations based on the missing: Towards contrastive explanations with perti- nent negatives. In: Bengio, S., Wallach, H.M., Larochelle, H., Grauman, K., Cesa- Bianchi, N., Garnett, R. (eds.) Advances in Neural Information Processing Systems 31: Annual Conference on Neural Inf...

    work page 2018

  8. [8]

    Doshi-Velez,F.,Kim,B.:Towardsarigorousscienceofinterpretablemachinelearn- ing (2017), https://arxiv.org/abs/1702.08608

    arXiv 2017

Show all 30 references
  1. [9]

    arXiv preprint arXiv:2405.01661 (2024), https://arxiv.org/abs/2405.01661

    Finzel, B., Hilme, P., Rabold, J., Schmid, U.: Telling more with concepts and relations: Exploring and evaluating classifier decisions with CoReX. arXiv preprint arXiv:2405.01661 (2024), https://arxiv.org/abs/2405.01661

    2024 arXiv

  2. [10]

    In: Julián, V., Ca- macho, D., Yin, H., Alberola, J.M., Nogueira, V.B., Novais, P., Tallón-Ballesteros, A.J

    Finzel, B., Knoblach, J., Thaler, A.M., Schmid, U.: Near hit and near miss example explanations for model revision in binary image classification. In: Julián, V., Ca- macho, D., Yin, H., Alberola, J.M., Nogueira, V.B., Novais, P., Tallón-Ballesteros, A.J. (eds.) Intelligent Da...

    2024

  3. [11]

    Fisher, R.A.: Statistical Methods for Research Workers, pp. 66–70. Springer New York, New York, NY (1992). https://doi.org/10.1007/978-1-4612-4380-9\_6

    1992 doi

  4. [12]

    Gunning, D., Aha, D.W.: Darpa’s explainable artificial intelligence (XAI) program. AI Mag. 40(2), 44–58 (2019). https://doi.org/10.1609/AIMAG.V40I2.2850

    2019 doi

  5. [13]

    Hernández-Orallo, J.: Gazing into clever hans machines. Nat. Mach. Intell.1(4), 172–173 (2019). https://doi.org/10.1038/S42256-019-0032-5 14 Y. Kaidashova et al

    2019 doi

  6. [14]

    In: Kelleher, C., Burnett, M.M., Sauer, S

    Kulesza, T., Stumpf, S., Burnett, M.M., Yang, S., Kwan, I., Wong, W.: Too much, too little, or just right? ways explanations impact end users’ men- tal models. In: Kelleher, C., Burnett, M.M., Sauer, S. (eds.) 2013 IEEE Symposium on Visual Languages and Human Centric Computing...

    2013

  7. [15]

    In: Proc

    vanLent,M.,Fisher,W.,Mancuso,M.:Anexplainableartificialintelligencesystem for small-unit tactical behavior. In: Proc. 2004 Nat. Conf. Artificial Intelligence. pp. 900–907. AAAI Press; MIT Press, 2004

    2004

  8. [16]

    Royal Institute of Philosophy Supplement27, 247–266 (1990)

    Lipton, P.: Contrastive explanation. Royal Institute of Philosophy Supplement27, 247–266 (1990). https://doi.org/10.1017/S1358246100005130

    1990 doi

  9. [17]

    https://doi.org/https://doi.org/10.1111/j.1468-0068.2007.00663.x

    Margolis, E., Laurence, S.: The ontology of concepts—abstract objects or mental representations? Noûs 41(4), 561–593 (2007). https://doi.org/https://doi.org/10.1111/j.1468-0068.2007.00663.x

    2007

  10. [18]

    Miller, G.A.: WordNet: A lexical database for english. Commun. ACM38(11), 39–41 (1995). https://doi.org/10.1145/219717.219748

    1995

  11. [19]

    Miller, T.: Explanation in artificial intelligence: Insights from the social sciences. Artif. Intell. 267, 1–38 (2019). https://doi.org/10.1016/J.ARTINT.2018.07.007

    2019 doi

  12. [20]

    Margolis and S

    Palmer, D.C.: Psychological essentialism: A review of E. Margolis and S. Laurence (eds.), Concepts: Core readings (2002)

    2002

  13. [21]

    Poeta, E., Ciravegna, G., Pastor, E., Cerquitelli, T., Baralis, E.: Concept-based ex- plainable artificial intelligence: A survey (2023), https://arxiv.org/abs/2312.12936

    2023

  14. [22]

    In: 2020 IEEE International Conference on Image Processing (ICIP)

    Prabhushankar, M., Kwon, G., Temel, D., AlRegib, G.: Contrastive explanations in neural networks. In: 2020 IEEE International Conference on Image Processing (ICIP). pp. 3289–3293. IEEE (2020)

    2020

  15. [23]

    Machine Learning111(5), 1799–1820 (2022)

    Rabold, J., Siebers, M., Schmid, U.: Generating contrastive explanations for induc- tive logic programming based on a near miss approach. Machine Learning111(5), 1799–1820 (2022). https://doi.org/10.1007/s10994-021-06048-w

    2022 doi

  16. [24]

    Rudin, C.: Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead. Nat. Mach. Intell.1(5), 206–215 (2019). https://doi.org/10.1038/S42256-019-0048-X

    2019 doi

  17. [25]

    International Journal of Computer Vision (IJCV) 115(3), 211–252 (2015)

    Russakovsky, O., Deng, J., Su, H., Krause, J., Satheesh, S., Ma, S., Huang, Z., Karpathy, A., Khosla, A., Bernstein, M., Berg, A.C., Fei-Fei, L.: ImageNet Large Scale Visual Recognition Challenge. International Journal of Computer Vision (IJCV) 115(3), 211–252 (2015). https://...

    2015 doi

  18. [26]

    Data Min

    Schwalbe, G., Finzel, B.: A comprehensive taxonomy for explainable artificial intel- ligence: a systematic survey of surveys on methods and concepts. Data Min. Knowl. Discov. 38(5), 3043–3101 (2024). https://doi.org/10.1007/S10618-022-00867-8

    2024 doi

  19. [27]

    In: Bengio, Y., LeCun, Y

    Simonyan, K., Zisserman, A.: Very deep convolutional networks for large-scale image recognition. In: Bengio, Y., LeCun, Y. (eds.) 3rd International Conference on Learning Representations, ICLR 2015, San Diego, CA, USA, May 7-9, 2015, Conference Track Proceedings (2015), http:/...

    2015 arXiv

  20. [28]

    IEEE Access 9, 11974–12001 (2021)

    Stepin, I., Alonso, J.M., Catalá, A., Pereira-Fariña, M.: A survey of contrastive and counterfactual explanation generation methods for ex- plainable artificial intelligence. IEEE Access 9, 11974–12001 (2021). https://doi.org/10.1109/ACCESS.2021.3051315

    2021

  21. [29]

    Biometrika 6(1), 1–25 (1908)

    Student: The probable error of a mean. Biometrika 6(1), 1–25 (1908). https://doi.org/10.2307/2331554

    1908 doi

  22. [30]

    In: The Psychology of Computer Vision, pp

    Winston, P.H.: Learning structural descriptions from examples. In: The Psychology of Computer Vision, pp. 157—-210. McGraw-Hill (1975) Toward Simple and Robust Contrastive Explanations 15 A Appendix 16 Y. Kaidashova et al. (a) Very strong relevance (b) Strong relevance (c) Low...

    1975

Pith tools

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