Pith. sign in

REVIEW 4 major objections 6 minor 41 references

MCCE: Missingness-aware Causal Concept Explainer

T0 review · 4 major / 6 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read The paper claims that missing concept annotations bias causal concept effect estimates, and that orthogonalizing a hidden representation against observed concepts removes most of that bias.

desk verdict The paper targets a real gap—missing concepts in causal concept effect estimation—but its pseudo-concept mechanism is mathematically inert: adding orthogonalized residuals does not change the observed-concept coefficients, so the claimed bias correction does not happen. read the letter →

arxiv 2411.09639 v1 pith:KN3SJQNC submitted 2024-11-14 cs.LG

Jifan Gao , Guanhua Chen This is my paper · ORCID
classification cs.LG
keywords causalconcepteffectconcept-basedexplainabilitymissingconceptspseudo-conceptsorthogonalprojectioninterpretablemachinelearningICaCECEBaB
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

Causal concept effect explanations estimate how much a human-interpretable concept moves a model's prediction, but existing methods assume every relevant concept is annotated. This paper shows that when concepts are missing, ordinary linear estimates of the observed concepts' causal effects are biased, so the resulting explanations are wrong. Its proposed MCCE builds pseudo-concepts, residuals of a learned representation after removing the projection onto the observed concepts, and adds them to a linear predictor of the black-box output. On the CEBaB restaurant-review benchmark, MCCE reports ICaCE-Error at or below the strongest baseline even when one or two of the four concepts are hidden, while needing only factual data. The paper also notes that validation is limited to CEBaB and that the number of pseudo-concepts is chosen empirically.

What carries the argument

The load-bearing object is the pseudo-concept matrix $C_\text{pseud} = (I-P)H$, where $H$ is the hidden representation extracted from the input, $P = C_\text{ob}(C_\text{ob}^T C_\text{ob})^{-1} C_\text{ob}^T$ is the least-squares projection matrix onto the observed concept columns, and $I$ is the identity. This is an orthogonalization: it removes from $H$ every direction already explainable by the observed concepts, leaving residual directions that are outside the observed concept span. The paper uses this residual as a proxy for the information that unobserved concepts would have contributed, and fits a linear model on the concatenated observed and pseudo concepts to approximate the black-box model's logits. The identity $C_\text{ob}^T C_\text{pseud} = 0$ is what prevents the pseudo-concepts from being a collinear duplicate of the observed concepts.

What would settle it

Construct a synthetic benchmark where a known unobserved concept is included in the true data-generating process, then explicitly remove all information about it from the feature extractor's hidden representation H (for example, by orthogonalizing H against the unobserved concept); if MCCE's ICaCE-Error on the observed concepts is then no better than an S-Learner that ignores missing concepts, the central claim that pseudo-concepts compensate for unobserved concepts is falsified.

Watch

Extended reading notes

Core claim

The central claim is that missing concept annotations need not be fatal. The authors analyze the linear concept-effect model $N(X) = C_\text{ob}^T \beta^*_\text{ob} + C_\text{un}^T \beta^*_\text{un}$ and observe that fitting only $C_\text{ob}$ leaves a residual $C_\text{un}^T \beta^*_\text{un} - C_\text{un}^T \hat{\beta}_\text{un}$ that moves the estimated $\hat{\beta}_\text{ob}$ away from $\beta^*_\text{ob}$; consequently the individual concept causal effect estimate is biased. They propose to recover the missing subspace information from a hidden representation $H$ by computing $C_\text{pseud} = (I-P)H$, with $P$ the projection onto the observed concept columns. Because $C_\text{ob}^T C_\text{pseud}=0$, the added pseudo-concepts cannot create collinearity with the observed concepts and, assuming $H$ contains the unobserved concepts' information, their inclusion compensates for the omitted variables. MCCE then trains a linear predictor on observed concepts plus pseudo-concepts and uses its coefficients to estimate causal effects both locally and globally. On CEBaB with BERT, RoBERTa, and Llama-3 extractors, the method's reported ICaCE-Error is lower than S-Learner in all reported settings and on par with or lower than the input-based Causal Proxy Model.

Load-bearing premise

The framework assumes the internal representation produced by the feature extractor still contains everything that would have been in the unobserved concepts; if it does not, the orthogonal residuals add nothing useful and the bias remains.

Editorial extensions

If this is right

  • If the hidden representation captures the unobserved concepts, MCCE removes the main obstacle to using concept-based causal explanation on partially annotated data.
  • MCCE requires only factual samples, whereas the strongest baseline it compares with needs labeled counterfactual texts.
  • The linear coefficients returned by MCCE give a direct global summary of how each concept affects the model's outputs.
  • The same pipeline can be used as an interpretable predictor, with macro-F1 comparable to the black-box model on CEBaB.

Reading between the lines

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

  • If the orthogonalization is doing the work, then deliberately removing unobserved-concept information from H (for example, by projecting H onto a subspace uncorrelated with the missing concept) should make MCCE's bias reduction disappear; this is a testable isolation of the mechanism.
  • The success of pseudo-concepts suggests a rank condition: the number of pseudo-concepts should roughly match the dimension of the unobserved-concept space, which would explain the paper's empirical observation that setting it comparable to or slightly larger than the number of observed concepts works best.
  • Because MCCE is representation-agnostic, the same missingness-aware correction could extend to image, tabular, or multimodal encoders, although the paper only demonstrates it on text.
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

4 major / 6 minor

Summary. The paper addresses causal concept effect estimation when some human-annotated concepts are missing. It first shows, using a linear model, that omitting unobserved concepts biases the estimated coefficients of observed concepts. It then proposes MCCE, which extracts a hidden representation H from the black-box model's input and constructs pseudo-concepts Cpseud = (I - P)H, where P projects onto the observed concepts. These pseudo-concepts are concatenated with the observed concepts in a linear predictor of the black-box output, and the resulting coefficients are used to estimate individual causal concept effects on the CEBaB dataset with BERT, RoBERTa, and Llama-3 backbones. The paper reports ICaCE-Error comparisons against approximate counterfactuals, S-Learner, and Causal Proxy Models, and also presents global interpretation coefficients and predictive performance of MCCE as an interpretable model.

Significance. If the proposed mechanism worked, MCCE would fill a real gap: current concept-based causal explainers assume complete concept annotations, which is unrealistic in practice. The paper's empirical setup is relevant: it uses a benchmark with human-verified counterfactual texts, evaluates multiple transformer backbones, and includes one- and two-concept missingness scenarios. The formal demonstration that omitted unobserved concepts bias observed-concept effect estimates is standard but clearly presented. However, the central claimed mechanism—that orthogonalized pseudo-concepts reduce this bias—is mathematically inert under the paper's own equations, and the empirical comparison does not isolate the effect of the pseudo-concepts. The paper also provides no theorem or formal condition under which (I - P)H captures the effect of unobserved concepts. Because the main contribution is not supported, the manuscript cannot be accepted in its current form.

major comments (4)
  1. [§3, Eq. (7)–(10)] The proposed pseudo-concepts cannot change the estimated coefficients of the observed concepts. Because Eq. (8) establishes C_ob^T Cpseud = 0, the normal equations for the joint OLS in Eq. (10) decouple, giving \hat\beta_ob = (C_ob^T C_ob)^{-1} C_ob^T N(X), which is exactly the estimate obtained by regressing N(X) on C_ob alone. Thus \hat\beta_ob is independent of H, of the number of pseudo-concepts j, and of \hat\beta_pseud. Consequently, the omitted-variable bias analyzed around Eq. (6) is unchanged, and by the paper's own Eq. (5) the ICaCE-Error for observed concepts is unchanged. The orthogonality property prevents collinearity but does not address the source of the bias, which is correlation between C_ob and C_un; any component of H carrying that correlated information is removed by the projection (I - P), while the retained orthogonal component is uncorrelated with C_ob and cannot adjust \hat\beta_ob. The paper's central claim that pseudo-concepts 'effectively mitigate the residue bias' is therefore not supported by the estimator defined in Eq. (10).
  2. [§3, Eq. (11)–(12)] The intervention formula E_MCCE(c, c'|x) is not a clean swap of the observed concept value while holding all other concepts fixed. In Eq. (11), the pseudo-concept term contains the projection matrix I - C_{x,ob}(C_{x,ob}^T C_{x,ob})^{-1} C_{x,ob}^T, which depends on the observed concept vector itself. Replacing C_{x,ob} by C_{x,ob}^{c→c'} therefore changes not only the observed-concept term but also the value of the pseudo-concept term, even though the raw input x is unchanged. The paper does not define what intervention is being applied to the pseudo-concepts or why this corresponds to the counterfactual in Eq. (4), where all remaining concepts are unchanged. This makes the causal interpretation of the reported estimates ambiguous and needs to be resolved for the main empirical claims to be meaningful.
  3. [§5, Table 1] The empirical comparison between MCCE and S-Learner does not demonstrate that pseudo-concepts reduce bias. S-Learner is a logistic regression, whereas MCCE uses a linear MSE predictor on hidden representations, so the comparison in Table 1 varies both the model class and the feature representation. A controlled ablation would compare a linear regression on the observed concepts alone against a linear regression on observed concepts plus pseudo-concepts, using the same extractor and training procedure. The paper instead calls S-Learner a 'special case of MCCE that removed the components of the pseudo-concepts,' which is not literally true because the two estimators have different loss functions and model classes. Given the mathematical inertness of the pseudo-concepts for \hat\beta_ob shown in Eq. (7)–(10), any observed ICaCE-Error improvement cannot be attributed to the proposed bias-correction mechanism without this ablation.
  4. [§3, text before Eq. (7)] No formal conditions are given under which the residual (I - P)H captures the effect of the unobserved concepts. The paper states, 'We hypothesize that H contains all necessary information about all concepts, including unobserved ones,' but this sufficiency assumption is never tested or formalized. If H encodes the unobserved concepts only along directions already collinear with the observed concepts, then (I - P)H contains no useful information about C_un. The paper needs either a theorem with explicit identification conditions or a direct diagnostic that the orthogonalized residuals carry information about the held-out concepts in the CEBaB experiments. As written, the theoretical contribution reduces to a demonstration of routine omitted-variable bias plus an untested representation-sufficiency assumption.
minor comments (6)
  1. [Eq. (10)] Equation (10) uses 'arg max' when minimizing a squared-error objective; it should be 'arg min.'
  2. [Eq. (7)] The phrase 'identification matrix' should read 'identity matrix.'
  3. [§2] There is a typo in 'summerizing' and another in 'Cconcept Credible Models'; both should be corrected.
  4. [§5, Figure 3] The sentence 'A positive Ambiance has a peaking impact at the 5-star rating instead of the 5-star rating' contains a repeated rating value and should be clarified.
  5. [§5, Table 2 paragraph] The acronym 'MCEE' appears where 'MCCE' is intended.
  6. [§5, first paragraph] The statement that S-Learner is 'a special case of MCCE that removed the components of the pseudo-concepts' is misleading because S-Learner uses logistic regression; please specify the linear-regression variant used as the ablation.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: MCCE is trained on factual black-box outputs and scored against external human-annotated counterfactuals; the orthogonal pseudo-concept construction is an explicit modeling assumption rather than a fitted prediction, though the bias-correction argument has a mathematical gap.

full rationale

The derivation is not circular. The target quantity, ICaCE-Error, is computed with CEBaB's human-written counterfactual inputs, which are not used to train MCCE; the explainer is fit to factual inputs and outputs, so the evaluation metric is external to the training objective. The pseudo-concepts are introduced as Cpseud=(I-P)H with P the projection onto observed concepts, and the paper openly states the hypothesis that H contains the needed unobserved-concept information; this is a modeling assumption, not an identity imported from prior work by the authors. No load-bearing self-citation appears; the only cited inspiration is Fan et al. (2024), an external factor-analysis reference. There is a genuine mathematical concern: because C_ob^T Cpseud=0, the normal equations leave the observed-concept OLS coefficient identical to a regression without pseudo-concepts, so Equations (7)-(10) do not themselves establish bias reduction for the observed-concept coefficients; and the inference-time projection in Equation (11) differs from the training-time projection in Equation (7). These are correctness/validity issues, not circular reductions of a prediction to its inputs, so under the circularity rubric the appropriate score is 0.

Assumptions & free parameters 1 free parameters · 3 assumptions · 1 invented entities

The central causal claim rests on three unproven assumptions: the existence of an exact linear explainer, the sufficiency of H to encode all concepts, and the sufficiency of the orthogonal residual to capture the missing-concept effect. The only reported hyperparameter, the number of pseudo-concepts, is chosen empirically without theoretical grounding. The paper introduces no external entities beyond the internal pseudo-concepts.

free parameters (1)
  • Number of pseudo-concepts j = Not reported for Table 1; stated optimal 'nine or twelve' in the one-unobserved scenario
    The count of pseudo-concepts is a hyperparameter selected by empirical performance. The paper acknowledges that the theoretical rationale is open, and it does not report the exact values used to produce the main results.
assumptions (3)
  • domain assumption There exists a linear explainer E* with coefficients beta* such that C_complete^T beta* = N(X) for the complete concept set.
    Invoked in Section 3, Equation (3). If the black-box model's logits are not linear in the concepts, the ICaCE is not equal to (c' - c) beta_t and the coefficient-based causal estimate is misspecified.
  • ad hoc to paper The hidden representation H contains all necessary information about all concepts, including unobserved ones.
    Stated as a hypothesis immediately before Equation (7). The entire bias-correction mechanism depends on H encoding C_un, but no evidence or argument is provided.
  • ad hoc to paper The residual (I - P)H after projecting H onto the observed-concept column space captures the effect of the unobserved concepts.
    Used in Equations (7) to (10). Even if H contains C_un, the unobserved concepts may have components in the span of C_ob or may not be linearly recoverable from H, so the residual is not guaranteed to match C_un's contribution.
invented entities (1)
  • Pseudo-concepts C_pseud = (I - P)H
    purpose: Latent variables intended to capture information lost by missing concepts and to reduce bias in the causal effect estimates.
    They are constructed from the extractor's hidden states and are not externally observable or independently falsifiable. The paper provides no handle outside the framework to verify that they correspond to unobserved concepts.

how reviews work

0 comments
Cite this review

Pith. "Pith review of MCCE: Missingness-aware Causal Concept Explainer." pith.science (2026). https://pith.science/paper/KN3SJQNC

@misc{pith2026241109639,
  author       = {Pith},
  title        = {Pith review of: MCCE: Missingness-aware Causal Concept Explainer},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KN3SJQNC}},
  note         = {Machine review of arXiv:2411.09639}
}
read the original abstract

Causal concept effect estimation is gaining increasing interest in the field of interpretable machine learning. This general approach explains the behaviors of machine learning models by estimating the causal effect of human-understandable concepts, which represent high-level knowledge more comprehensibly than raw inputs like tokens. However, existing causal concept effect explanation methods assume complete observation of all concepts involved within the dataset, which can fail in practice due to incomplete annotations or missing concept data. We theoretically demonstrate that unobserved concepts can bias the estimation of the causal effects of observed concepts. To address this limitation, we introduce the Missingness-aware Causal Concept Explainer (MCCE), a novel framework specifically designed to estimate causal concept effects when not all concepts are observable. Our framework learns to account for residual bias resulting from missing concepts and utilizes a linear predictor to model the relationships between these concepts and the outputs of black-box machine learning models. It can offer explanations on both local and global levels. We conduct validations using a real-world dataset, demonstrating that MCCE achieves promising performance compared to state-of-the-art explanation methods in causal concept effect estimation.

Figures

Figures reproduced from arXiv: 2411.09639 by the authors.

Figure 1
Figure 1. The architecture of MCCE. Given an input sample, a vector representation is extracted. [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Causal structure graph. The impact of U on X is not only mediated by the observed concepts Cob1 , ...Cobk but also by the unobserved concepts Cun1 , ...Cunj . In this work, we aim to account for the impact of unobserved concepts when estimating the causal effect of observed con￾cepts, which has not been addressed in existing research. A backdoor path may exist from Cob1 to N (X), even though there is no direct path … view at source ↗
Figure 3
Figure 3. An illustration of the MCCE’s global interpretation on a BERT model when the concepts [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

41 extracted references · 19 canonical work pages

  1. [1]

    Cebab: Estimating the causal effects of real-world concepts on nlp model behavior

    Eldar D Abraham, Karel D'Oosterlinck, Amir Feder, Yair Gat, Atticus Geiger, Christopher Potts, Roi Reichart, and Zhengxuan Wu. Cebab: Estimating the causal effects of real-world concepts on nlp model behavior. Advances in Neural Information Processing Systems, 35: 0 17582--17596, 2022

    work page 2022

  2. [2]

    Permutation importance: a corrected feature importance measure

    Andr \'e Altmann, Laura Tolo s i, Oliver Sander, and Thomas Lengauer. Permutation importance: a corrected feature importance measure. Bioinformatics, 26 0 (10): 0 1340--1347, 2010

    work page 2010

  3. [3]

    Logic explained networks

    Gabriele Ciravegna, Pietro Barbiero, Francesco Giannini, Marco Gori, Pietro Li \'o , Marco Maggini, and Stefano Melacci. Logic explained networks. Artificial Intelligence, 314: 0 103822, 2023

    work page 2023

  4. [4]

    Machine intelligence in healthcare—perspectives on trustworthiness, explainability, usability, and transparency

    Christine M Cutillo, Karlie R Sharma, Luca Foschini, Shinjini Kundu, Maxine Mackintosh, Kenneth D Mandl, and MI in Healthcare Workshop Working Group Beck Tyler 1 Collier Elaine 1 Colvis Christine 1 Gersing Kenneth 1 Gordon Valery 1 Jensen Roxanne 8 Shabestari Behrouz 9 Southall Noel 1. Machine intelligence in healthcare—perspectives on trustworthiness, ex...

    work page 2020

  5. [5]

    Bert: Pre-training of deep bidirectional transformers for language understanding

    Jacob Devlin. Bert: Pre-training of deep bidirectional transformers for language understanding. arXiv preprint arXiv:1810.04805, 2018

    arXiv 2018

  6. [6]

    The llama 3 herd of models

    Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Amy Yang, Angela Fan, et al. The llama 3 herd of models. arXiv preprint arXiv:2407.21783, 2024

    arXiv 2024

  7. [7]

    Are latent factor regression and sparse regression adequate? Journal of the American Statistical Association, 119 0 (546): 0 1076--1088, 2024

    Jianqing Fan, Zhipeng Lou, and Mengxin Yu. Are latent factor regression and sparse regression adequate? Journal of the American Statistical Association, 119 0 (546): 0 1076--1088, 2024

    work page 2024

  8. [8]

    Causalm: Causal model explanation through counterfactual language models

    Amir Feder, Nadav Oved, Uri Shalit, and Roi Reichart. Causalm: Causal model explanation through counterfactual language models. Computational Linguistics, 47 0 (2): 0 333--386, 2021

    2021

Show all 41 references
  1. [9]

    Towards automatic concept-based explanations

    Amirata Ghorbani, James Wexler, James Y Zou, and Been Kim. Towards automatic concept-based explanations. Advances in neural information processing systems, 32, 2019

    2019

  2. [10]

    Safe artificial intelligence in finance

    Paolo Giudici and Emanuela Raffinetti. Safe artificial intelligence in finance. Finance Research Letters, 56: 0 104088, 2023

    2023

  3. [11]

    Distilling the knowledge in a neural network

    Geoffrey Hinton. Distilling the knowledge in a neural network. arXiv preprint arXiv:1503.02531, 2015

    2015 arXiv

  4. [12]

    Interpretability beyond feature attribution: Quantitative testing with concept activation vectors (tcav)

    Been Kim, Martin Wattenberg, Justin Gilmer, Carrie Cai, James Wexler, Fernanda Viegas, et al. Interpretability beyond feature attribution: Quantitative testing with concept activation vectors (tcav). In International conference on machine learning, pp.\ 2668--2677. PMLR, 2018

    2018

  5. [13]

    Probabilistic concept bottleneck models

    Eunji Kim, Dahuin Jung, Sangha Park, Siwon Kim, and Sungroh Yoon. Probabilistic concept bottleneck models. arXiv preprint arXiv:2306.01574, 2023

    2023 arXiv

  6. [14]

    Concept bottleneck models

    Pang Wei Koh, Thao Nguyen, Yew Siang Tang, Stephen Mussmann, Emma Pierson, Been Kim, and Percy Liang. Concept bottleneck models. In International conference on machine learning, pp.\ 5338--5348. PMLR, 2020

    2020

  7. [15]

    o ren R K \

    S \"o ren R K \"u nzel, Jasjeet S Sekhon, Peter J Bickel, and Bin Yu. Metalearners for estimating heterogeneous treatment effects using machine learning. Proceedings of the national academy of sciences, 116 0 (10): 0 4156--4165, 2019

    2019

  8. [16]

    Roberta: A robustly optimized bert pretraining approach

    Yinhan Liu. Roberta: A robustly optimized bert pretraining approach. arXiv preprint arXiv:1907.11692, 2019

    1907 arXiv

  9. [17]

    A unified approach to interpreting model predictions

    Scott Lundberg. A unified approach to interpreting model predictions. arXiv preprint arXiv:1705.07874, 2017

    2017 arXiv

  10. [18]

    Interpretable machine learning

    Christoph Molnar. Interpretable machine learning. Lulu. com, 2020

    2020

  11. [19]

    Causal interpretability for machine learning-problems, methods and evaluation

    Raha Moraffah, Mansooreh Karami, Ruocheng Guo, Adrienne Raglin, and Huan Liu. Causal interpretability for machine learning-problems, methods and evaluation. ACM SIGKDD Explorations Newsletter, 22 0 (1): 0 18--33, 2020

    2020

  12. [20]

    black box

    Julian D Olden and Donald A Jackson. Illuminating the “black box”: a randomization approach for understanding variable contributions in artificial neural networks. Ecological modelling, 154 0 (1-2): 0 135--150, 2002

    2002

  13. [21]

    Causality

    J Pearl. Causality. Cambridge university press, 2009

    2009

  14. [22]

    Concept-based explainable artificial intelligence: A survey

    Eleonora Poeta, Gabriele Ciravegna, Eliana Pastor, Tania Cerquitelli, and Elena Baralis. Concept-based explainable artificial intelligence: A survey. arXiv preprint arXiv:2312.12936, 2023

    2023

  15. [23]

    Explainable, trustworthy, and ethical machine learning for healthcare: A survey

    Khansa Rasheed, Adnan Qayyum, Mohammed Ghaly, Ala Al-Fuqaha, Adeel Razi, and Junaid Qadir. Explainable, trustworthy, and ethical machine learning for healthcare: A survey. Computers in Biology and Medicine, 149: 0 106043, 2022

    2022

  16. [24]

    Null it out: Guarding protected attributes by iterative nullspace projection

    Shauli Ravfogel, Yanai Elazar, Hila Gonen, Michael Twiton, and Yoav Goldberg. Null it out: Guarding protected attributes by iterative nullspace projection. arXiv preprint arXiv:2004.07667, 2020

    2004 arXiv

  17. [25]

    Attention-based interpretability with concept transformers

    Mattia Rigotti, Christoph Miksovic, Ioana Giurgiu, Thomas Gschwind, and Paolo Scotton. Attention-based interpretability with concept transformers. In International conference on learning representations, 2021

    2021

  18. [26]

    Model class reliance for random forests

    Gavin Smith, Roberto Mansilla, and James Goulding. Model class reliance for random forests. Advances in Neural Information Processing Systems, 33: 0 22305--22315, 2020

    2020

  19. [27]

    Rethinking the role of gradient-based attribution methods for model interpretability

    Suraj Srinivas and Fran c ois Fleuret. Rethinking the role of gradient-based attribution methods for model interpretability. arXiv preprint arXiv:2006.09128, 2020

    2006 arXiv

  20. [28]

    Axiomatic attribution for deep networks

    Mukund Sundararajan, Ankur Taly, and Qiqi Yan. Axiomatic attribution for deep networks. In International conference on machine learning, pp.\ 3319--3328. PMLR, 2017

    2017

  21. [29]

    Sparsity-guided holistic explanation for llms with interpretable inference-time intervention

    Zhen Tan, Tianlong Chen, Zhenyu Zhang, and Huan Liu. Sparsity-guided holistic explanation for llms with interpretable inference-time intervention. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 38, pp.\ 21619--21627, 2024

    2024

  22. [30]

    Learning concept credible models for mitigating shortcuts

    Jiaxuan Wang, Sarah Jabbour, Maggie Makar, Michael Sjoding, and Jenna Wiens. Learning concept credible models for mitigating shortcuts. Advances in neural information processing systems, 35: 0 33343--33356, 2022

    2022

  23. [31]

    Causal proxy models for concept-based model explanations

    Zhengxuan Wu, Karel D’Oosterlinck, Atticus Geiger, Amir Zur, and Christopher Potts. Causal proxy models for concept-based model explanations. In International conference on machine learning, pp.\ 37313--37334. PMLR, 2023

    2023

  24. [32]

    Energy-based concept bottleneck models: Unifying prediction, concept intervention, and probabilistic interpretations

    Xinyue Xu, Yi Qin, Lu Mi, Hao Wang, and Xiaomeng Li. Energy-based concept bottleneck models: Unifying prediction, concept intervention, and probabilistic interpretations. In The Twelfth International Conference on Learning Representations, 2024

    2024

  25. [33]

    On completeness-aware concept-based explanations in deep neural networks

    Chih-Kuan Yeh, Been Kim, Sercan Arik, Chun-Liang Li, Tomas Pfister, and Pradeep Ravikumar. On completeness-aware concept-based explanations in deep neural networks. Advances in neural information processing systems, 33: 0 20554--20565, 2020

    2020

  26. [34]

    Post-hoc concept bottleneck models

    Mert Yuksekgonul, Maggie Wang, and James Zou. Post-hoc concept bottleneck models. arXiv preprint arXiv:2205.15480, 2022

    2022 arXiv

  27. [35]

    Concept embedding models: Beyond the accuracy-explainability trade-off

    Mateo Espinosa Zarlenga, Pietro Barbiero, Gabriele Ciravegna, Giuseppe Marra, Francesco Giannini, Michelangelo Diligenti, Zohreh Shams, Frederic Precioso, Stefano Melacci, Adrian Weller, et al. Concept embedding models: Beyond the accuracy-explainability trade-off. arXiv prepr...

    2022 arXiv

  28. [36]

    Learning deep features for discriminative localization

    Bolei Zhou, Aditya Khosla, Agata Lapedriza, Aude Oliva, and Antonio Torralba. Learning deep features for discriminative localization. In Proceedings of the IEEE conference on computer vision and pattern recognition, pp.\ 2921--2929, 2016

    2016

  29. [37]

    Finbrain 2.0: when finance meets trustworthy ai

    Jun Zhou, Chaochao Chen, Longfei Li, Zhiqiang Zhang, and Xiaolin Zheng. Finbrain 2.0: when finance meets trustworthy ai. Frontiers of Information Technology & Electronic Engineering, 23 0 (12): 0 1747--1764, 2022

    2022

  30. [38]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

  31. [39]

    @esa (Ref

    \@ifxundefined[1] #1\@undefined \@firstoftwo \@secondoftwo \@ifnum[1] #1 \@firstoftwo \@secondoftwo \@ifx[1] #1 \@firstoftwo \@secondoftwo [2] @ #1 \@temptokena #2 #1 @ \@temptokena \@ifclassloaded agu2001 natbib The agu2001 class already includes natbib coding, so you should ...

  32. [40]

    \@lbibitem[] @bibitem@first@sw\@secondoftwo \@lbibitem[#1]#2 \@extra@b@citeb \@ifundefined br@#2\@extra@b@citeb \@namedef br@#2 \@nameuse br@#2\@extra@b@citeb \@ifundefined b@#2\@extra@b@citeb @num @parse #2 @tmp #1 NAT@b@open@#2 NAT@b@shut@#2 \@ifnum @merge>\@ne @bibitem@firs...

  33. [41]

    @open @close @open @close and [1] URL: #1 \@ifundefined chapter * \@mkboth \@ifxundefined @sectionbib * \@mkboth * \@mkboth\@gobbletwo \@ifclassloaded amsart * \@ifclassloaded amsbook * \@ifxundefined @heading @heading NAT@ctr thebibliography [1] @ \@biblabel @NAT@ctr \@bibset...

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.