Pith. sign in

REVIEW 4 major objections 3 minor 45 references

Multi-Rationale Explainable Object Recognition via Contrastive Conditional Inference

T0 review · 4 major / 3 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read The abstract claims that training-free contrastive rationale conditioning achieves state-of-the-art explainable object recognition.

desk verdict The submission is a mismatched artifact: the title/abstract promise an explainable object recognition paper, but the body is an unrelated condensed-matter preprint, so there is no supporting evidence for any of the advertised claims. read the letter →

arxiv 2508.14280 v1 pith:M7M2MWYU submitted 2025-08-19 cs.CV

Ali Rasekh , Sepehr Kazemi Ranjbar , Simon Gottschalk This is my paper · ORCID
classification cs.CV
keywords explainableobjectrecognitionvision-languagemodelsCLIPcontrastiveconditionalinferencemulti-rationalebenchmarkzero-shotclassificationrationalequalitytraining-free
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, as announced in its abstract, sets out to make explainable object recognition both more accurate and more convincingly explained by conditioning a vision-language model on multiple rationales per image instead of a single prompt. It introduces a multi-rationale benchmark (several ground-truth rationales for each image, plus evaluation metrics for rationale quality) and a contrastive conditional inference (CCI) framework that models the probabilistic relationships among image embeddings, category labels, and rationales. The headline claim is that CCI needs no training and still achieves state-of-the-art classification and strong zero-shot performance on this benchmark. If true, multiple structured rationales could sharpen CLIP-based zero-shot prediction while giving a fuller account of why a category was chosen. However, the supplied full text is an unrelated condensed-matter physics preprint, so the benchmark, the CCI equations, and the experiments described in the abstract do not appear anywhere in the document.

What carries the argument

Contrastive conditional inference (CCI): a parameter-free conditioning rule that combines CLIP's image embedding with several rationale text embeddings through contrastive probabilistic comparisons, replacing single-prompt conditioning. It is the mechanism by which multiple rationales are supposed to constrain the predicted category without training. The other load-bearing piece is the multi-rationale benchmark and its rationale-quality metrics, which are meant to make 'explanation quality' an objective evaluation axis rather than a qualitative bonus.

What would settle it

Take the benchmark images and run CCI with genuine multi-rationale annotations; then rerun the same classifier with the rationales randomly swapped among images. If classification accuracy does not drop, the rationales are not conditioning the prediction, and the claimed mechanism fails.

Watch

Extended reading notes

Core claim

The intended contribution is a training-free inference procedure for CLIP-style models: instead of asking the text encoder to absorb rationale structure through prompt text, CCI explicitly writes the relationship among image, category, and rationales as a joint conditional model and uses contrastive comparisons among rationale embeddings to infer the category. The paper also contributes a benchmark where each image carries multiple ground-truth rationales and metrics that score both class accuracy and rationale quality. The claimed discovery is that this multi-rationale contrastive conditioning improves classification over earlier prompt-based conditioning and, in zero-shot settings, sets a

Load-bearing premise

The load-bearing premise is that several rationale text embeddings, combined with the image embedding by contrastive probabilities and no learned weights, provide non-redundant evidence that improves category prediction over CLIP's plain zero-shot score, and that the new benchmark's ground-truth rationales and metrics are a valid measure of explanation quality.

Editorial extensions

If this is right

  • If CCI works as claimed, CLIP-based explainable recognition can improve without fine-tuning, since the inference is training-free.
  • Benchmarks with several ground-truth rationales per image would make explanation evaluation more reliable than single, noisy rationales.
  • Strong zero-shot performance would let the method deploy immediately on new categories without per-dataset adaptation.
  • A new state-of-the-art baseline would force future explainable-recognition work to report both accuracy and rationale quality on the same protocol.

Reading between the lines

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

  • Because the supplied full text is an unrelated physics preprint, the method and benchmark sections promised by the abstract are not present; the result claims should be treated as unverified until the actual computer-vision manuscript is available.
  • The pivotal empirical test is whether multiple rationales are non-redundant: if combining two rationales yields the same accuracy as one, the conditional-inference machinery is not doing extra work over prompt ensembling.
  • A testable extension is to measure per-rationale fidelity with human raters, for example whether the rationales that CCI actually uses to shift a prediction are the ones a human would judge as decisive.
  • The claimed zero-shot success would suggest that CLIP's embedding space already contains enough structured 'why' information to support explanation, which could motivate probing which layers or features encode rationale knowledge.
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 / 3 minor

Summary. The manuscript titled 'Multi-Rationale Explainable Object Recognition via Contrastive Conditional Inference' is submitted under arXiv:2508.14280 (cs.CV) and its abstract claims a new multi-rationale explainable object recognition benchmark, a training-free contrastive conditional inference (CCI) framework, and state-of-the-art accuracy and rationale quality. However, the submitted full text is an entirely different, unrelated condensed-matter manuscript, arXiv:2508.14283v2, titled 'Role of electron-electron interactions in M-valley twisted transition metal dichalcogenides.' The body, figures, tables, and Supplemental Material contain Hartree-HF calculations for moiré TMDs, with no mention of CLIP, rationales, object recognition, benchmarks, CCI, or any of the claimed experiments. The document therefore provides no support for the abstract's central claims.

Significance. If the abstract's claims were substantiated, a training-free contrastive conditional inference framework over CLIP embeddings could be a meaningful contribution to explainable vision-language classification. However, this submission provides no evidence whatsoever. There is no benchmark description, no metric definition, no CCI equations, no algorithm, no experiments, no code, and no machine-checked proofs. The claimed 'state-of-the-art results' and 'new standard' are entirely unsupported. The only technical content in the full text concerns electron-electron interactions in M-valley twisted TMDs, which is irrelevant to the stated computer-vision topic. The potential significance of the claimed work cannot be assessed from this document.

major comments (4)
  1. [Full text (all sections)] The submitted full text from 'Role of electron-electron interactions...' through the Supplemental Material is a different manuscript, identified by the footer 'arXiv:2508.14283v2 [cond-mat.mes-hall] 22 Sep 2025.' There is no occurrence of 'contrastive conditional inference,' 'CLIP,' 'rationale,' 'benchmark,' or 'object recognition' in the body. Thus the abstract's central claim of a CCI framework and benchmark has no in-text support whatsoever.
  2. [Abstract / benchmark construction] The abstract asserts a 'multi-rationale explainable object recognition benchmark comprising datasets in which each image is annotated with multiple ground-truth rationales' and evaluation metrics. The document identifies no datasets, no annotation protocol, no image sources, no number of rationales per image, and no definitions of the claimed evaluation metrics. Since the reported state-of-the-art is on this benchmark, the absence of the benchmark is load-bearing.
  3. [Abstract / CCI framework] The proposed 'contrastive conditional inference (CCI) framework' is never defined. There is no probabilistic model linking image embeddings, category labels, and rationales; no contrastive objective; no equations; and no algorithm. The claim that this framework works 'without requiring any training' is vacuous in this document. In particular, the load-bearing premise that training-free combination of rationale embeddings with the image embedding improves category prediction over CLIP's zero-shot score is neither derived nor tested.
  4. [All figures and tables] Every reported result in the submitted document is a Hartree-Fock band-structure, density-of-states, phase-diagram, or fitting-parameter result for twisted TMDs (Figs. 1–9, Tables S1, Figs. S1–S6). These results cannot substantiate any claim about object recognition accuracy or rationale quality. They are unrelated to the abstract and to the journal's stated topic.
minor comments (3)
  1. [Title / Abstract] The title and abstract do not match the full text. The abstract describes a computer-vision paper, while the body is a condensed-matter physics preprint. The editorial office should flag this as a metadata or file mismatch.
  2. [References] The reference list contains no citations to CLIP, vision-language models, explainable AI, or object recognition benchmarks; it is appropriate to the TMD physics paper, not to the claimed framework.
  3. [arXiv footer] The footer 'arXiv:2508.14283v2 [cond-mat.mes-hall] 22 Sep 2025' identifies the body as a different preprint. This self-identification should be explicitly resolved before any further review.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity can be established because the submitted full text contains none of the abstract's claimed method, benchmark, or experiments; the problem is missing support, not circular reduction.

full rationale

The abstract promises a multi-rationale explainable object recognition benchmark, a contrastive conditional inference (CCI) framework, and state-of-the-art results, but the full text is an unrelated condensed-matter preprint titled "Role of electron-electron interactions in M-valley twisted transition metal dichalcogenides" with footer "arXiv:2508.14283v2 [cond-mat.mes-hall] 22 Sep 2025". There is no CCI model, no benchmark dataset, no metric definition, no CLIP-based method, and no experiment in the document. A circularity finding requires quoting an equation or construction that reduces a prediction to its own inputs, e.g. a fitted parameter renamed as a prediction, or a definition that builds in the target result. No such reduction is present because the relevant derivation chain is absent entirely. The abstract's "state-of-the-art results on the multi-rationale explainable object recognition benchmark" is an unsupported assertion and a potential self-evaluation concern because the benchmark is introduced by the same work, but without the benchmark's metric definitions and experimental protocol there is no exhibited identity between the claimed prediction and the method's inputs. I therefore flag this as a completeness/verifiability failure rather than circularity; per the audit rules, absence of evidence is not circularity and should not inflate the circularity score.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

This ledger audits the only content present, the condensed-matter body, where the central quantities are built on a DFT-fitted continuum model and a phenomenological screening parameter (epsilon = 25, d = 20 nm). The ledger is therefore about a different paper than the abstract advertises; the abstract's claimed benchmark and CCI framework have no free parameters, axioms, or entities assignable because they are never described. The mismatch itself is the decisive observation for the verdict.

free parameters (4)
  • Moire continuum model parameters (mass, moire potentials, tunneling amplitudes and phases) = Table S1: e.g. aligned 3.89 deg: m_x=0.21, m_y=0.73, v11=29.4 meV, w1=110.87 meV, w2=-7.99 meV
    Symmetry-constrained fits to DFT taken from Ref. [23]; these fix the noninteracting Hamiltonian (Eqs. (1)-(3)) that all interaction results inherit.
  • Relative dielectric constant epsilon = epsilon = 25
    Phenomenological knob for interaction strength in U_q = e^2 tanh(qd)/(2 epsilon epsilon0 q); the paper states epsilon about 25 is the physically relevant scenario because it reproduces experimental features of other TMD moire systems (2H-tWSe2).
  • Gate-to-sample distance d = 20 nm
    Dual-gate screening length set by hand; results (bandwidth, DOS, phase diagram) are evaluated at this value.
  • HF band-basis cutoff N = N = 3 (6N N_F Green's function dimension)
    Truncation of the band basis in the Hartree-Fock iteration; the authors state that increasing the cutoff does not change results qualitatively, but the main results are computed at this cutoff.
assumptions (5)
  • domain assumption The single-particle moire continuum Hamiltonian (Eqs. (1)-(3)) with parameters fitted to DFT is a faithful low-energy model of 1T tSnSe2.
    Section 'Model' and Table S1: the model is 'previously constructed and studied in Refs. [22,23]' (the papers that proposed the M-valley platform); all interaction results inherit this model without independent verification in this work.
  • domain assumption Self-consistent Hartree-Fock captures the relevant weak-coupling physics.
    The paper studies the weak-coupling regime via HF and cites Refs. [28,29] for the framework; no comparison with other many-body methods (e.g., exact diagonalization, DMRG) is given in the main text.
  • domain assumption Approximate D3 symmetry and inherited monolayer SU(2) spin symmetry survive twisting.
    Section 'Model': emergent point group D3 generated by C3z and C2x (or C2y) for aligned (antialigned) stacking; the low-energy theory assumes these symmetries and approximates spin degeneracy.
  • domain assumption Local-stacking approximation holds for the moire potentials.
    Section 'Model': the potentials obey the local-stacking approximation for aligned stacking 'while for antialigned stacking, this is only approximately true'; the paper proceeds with it in both cases.
  • domain assumption Intervalley-coherent order can be excluded at half filling.
    The half-filling analysis restricts density matrices to be diagonal in valley space; the text states 'this analysis excludes intervalley-coherent order', while later mentioning fluctuations of a nascent intervalley coherent state as pairing glue, a tension in the discussion.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Multi-Rationale Explainable Object Recognition via Contrastive Conditional Inference." pith.science (2026). https://pith.science/paper/M7M2MWYU

@misc{pith2026250814280,
  author       = {Pith},
  title        = {Pith review of: Multi-Rationale Explainable Object Recognition via Contrastive Conditional Inference},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/M7M2MWYU}},
  note         = {Machine review of arXiv:2508.14280}
}
read the original abstract

Explainable object recognition using vision-language models such as CLIP involves predicting accurate category labels supported by rationales that justify the decision-making process. Existing methods typically rely on prompt-based conditioning, which suffers from limitations in CLIP's text encoder and provides weak conditioning on explanatory structures. Additionally, prior datasets are often restricted to single, and frequently noisy, rationales that fail to capture the full diversity of discriminative image features. In this work, we introduce a multi-rationale explainable object recognition benchmark comprising datasets in which each image is annotated with multiple ground-truth rationales, along with evaluation metrics designed to offer a more comprehensive representation of the task. To overcome the limitations of previous approaches, we propose a contrastive conditional inference (CCI) framework that explicitly models the probabilistic relationships among image embeddings, category labels, and rationales. Without requiring any training, our framework enables more effective conditioning on rationales to predict accurate object categories. Our approach achieves state-of-the-art results on the multi-rationale explainable object recognition benchmark, including strong zero-shot performance, and sets a new standard for both classification accuracy and rationale quality. Together with the benchmark, this work provides a more complete framework for evaluating future models in explainable object recognition. The code will be made available online.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

45 extracted references · 35 canonical work pages

  1. [1]

    CDUL: CLIP-Driven Unsupervised Learning for Multi-Label Image Classification

    Rabab Abdelfattah, Qing Guo, Xiaoguang Li, Xiaofeng Wang, and Song Wang. CDUL: CLIP-Driven Unsupervised Learning for Multi-Label Image Classification . In Proceedings of the IEEE/CVF international conference on computer vision, pages 1348--1357, 2023

    work page 2023

  2. [2]

    Multimodal video retrieval with CLIP: a user study

    Tayfun Alpay, Sven Magg, Philipp Broze, and Daniel Speck. Multimodal video retrieval with CLIP: a user study . Information Retrieval Journal, 26 0 (1): 0 6, 2023

    work page 2023

  3. [3]

    Interpreting CLIP with Sparse Linear Concept Embeddings (SpLiCE)

    Usha Bhalla, Alex Oesterling, Suraj Srinivas, Flavio Calmon, and Himabindu Lakkaraju. Interpreting CLIP with Sparse Linear Concept Embeddings (SpLiCE) . Advances in Neural Information Processing Systems, 37: 0 84298--84328, 2024

    work page 2024

  4. [4]

    Food-101--Mining Discriminative Components with Random Forests

    Lukas Bossard, Matthieu Guillaumin, and Luc Van Gool. Food-101--Mining Discriminative Components with Random Forests . In Computer vision--ECCV 2014: 13th European conference, zurich, Switzerland, September 6-12, 2014, proceedings, part VI 13, pages 446--461. Springer, 2014

    work page 2014

  5. [5]

    Grad-CAM++: Generalized Gradient-based Visual Explanations for Deep Convolutional Networks

    Aditya Chattopadhay, Anirban Sarkar, Pikee Singh, and Vineeth N Oza. Grad-CAM++: Generalized Gradient-based Visual Explanations for Deep Convolutional Networks . arXiv preprint arXiv:1710.11063, 2018

    arXiv 2018

  6. [6]

    Generic Attention-Model Explainability for Interpreting Bi-Modal and Encoder-Decoder Transformers

    Hila Chefer, Shaked Gur, and Lior Wolf. Generic Attention-Model Explainability for Interpreting Bi-Modal and Encoder-Decoder Transformers . In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), pages 3841--3850, 2021

    work page 2021

  7. [7]

    This looks like that: Deep learning for interpretable image recognition

    Chaofan Chen, Alvin Li, Chao Tao, Mayank Barnett, Jonathan Su, Yinqing Li, Madeline Carspecken, Akash Mukherjee, Kate Di Mauro, Tian intensifying Wang, et al. This looks like that: Deep learning for interpretable image recognition . In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 2892--2901, 2019

    work page 2019

  8. [8]

    gScoreCAM: What objects is CLIP looking at? In Proceedings of the Asian Conference on Computer Vision, pages 1959--1975, 2022

    Peijie Chen, Qi Li, Saad Biaz, Trung Bui, and Anh Nguyen. gScoreCAM: What objects is CLIP looking at? In Proceedings of the Asian Conference on Computer Vision, pages 1959--1975, 2022

    work page 1959

Show all 45 references
  1. [9]

    On the Difference of BERT-style and CLIP-style Text Encoders

    Zhihong Chen, Guiming Hardy Chen, Shizhe Diao, Xiang Wan, and Benyou Wang. On the Difference of BERT-style and CLIP-style Text Encoders . arXiv preprint arXiv:2306.03678, 2023

    2023 arXiv

  2. [10]

    ImageNet: A large-scale hierarchical image database

    Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. ImageNet: A large-scale hierarchical image database . In 2009 IEEE Conference on Computer Vision and Pattern Recognition, pages 248--255, 2009. doi:10.1109/CVPR.2009.5206848

    2009

  3. [11]

    Vision-Language Models for Autonomous Driving: CLIP-Based Dynamic Scene Understanding

    Mohammed Elhenawy, Huthaifa I Ashqar, Andry Rakotonirainy, Taqwa I Alhadidi, Ahmed Jaber, and Mohammad Abu Tami. Vision-Language Models for Autonomous Driving: CLIP-Based Dynamic Scene Understanding . Electronics, 14 0 (7): 0 1282, 2025

    2025

  4. [12]

    On Calibration of Modern Neural Networks

    Chuan Guo, Geoff Pleiss, Yu Sun, and Kilian Q Weinberger. On Calibration of Modern Neural Networks . In International conference on machine learning, pages 1321--1330. PMLR, 2017

    2017

  5. [13]

    The uncertainty and explainability in object recognition

    Wei Hui and Liping Yu. The uncertainty and explainability in object recognition . Journal of Experimental & Theoretical Artificial Intelligence, pages 1--20, 2020

    2020

  6. [14]

    Visual Prompt Tuning

    Menglin Jia, Luming Tang, Bor-Chun Chen, Claire Cardie, Serge Belongie, Bharath Hariharan, and Ser-Nam Lim. Visual Prompt Tuning . In European conference on computer vision, pages 709--727. Springer, 2022

    2022

  7. [15]

    Text encoders bottleneck compositionality in contrastive vision-language models

    Amita Kamath, Jack Hessel, and Kai-Wei Chang. Text encoders bottleneck compositionality in contrastive vision-language models . arXiv preprint arXiv:2305.14897, 2023

    2023 arXiv

  8. [16]

    E-ViL: A Dataset and Benchmark for Natural Language Explanations in Vision-Language Tasks

    Maxime Kayser, Virginie Do, Zeynep Akata, Leonard Salewski, Thomas Lukasiewicz, Oana-Maria Camburu, and Cornelius Emde. E-ViL: A Dataset and Benchmark for Natural Language Explanations in Vision-Language Tasks . In Proceedings of the IEEE/CVF International Conference on Comput...

    2021

  9. [17]

    Interpretability Beyond Feature Attribution: Quantitative Testing with Concept Activation Vectors ( TCAV )

    Been Kim, Martin Wattenberg, Justin Gilmer, Carrie Cai, James Wexler, Fernanda Viegas, and Rory sayres. Interpretability Beyond Feature Attribution: Quantitative Testing with Concept Activation Vectors ( TCAV ) . In Proceedings of the 35th International Conference on Machine L...

    2018

  10. [18]

    Concept Bottleneck Models

    Pang Wei Koh, Thao Nguyen, Yew Siang Tang, Stephen Mussmann, Emma Pierson, Been Kim, and Percy Liang. Concept Bottleneck Models . In Proceedings of the 37th International Conference on Machine Learning, volume 119, pages 5338--5348. PMLR, 13--18 Jul 2020

    2020

  11. [19]

    Learning multiple layers of features from tiny images

    Alex Krizhevsky. Learning multiple layers of features from tiny images. 2009. URL https://api.semanticscholar.org/CorpusID:18268744

    2009

  12. [20]

    Learning to detect unseen object classes by between-class attribute transfer

    Christoph H Lampert, Hannes Nickisch, and Stefan Harmeling. Learning to detect unseen object classes by between-class attribute transfer . In Computer Vision and Pattern Recognition, 2009. CVPR 2009. IEEE Conference on, pages 2413--2420. IEEE, 2009

    2009

  13. [21]

    Caltech 101 , Apr 2022

    Fei-Fei Li, Marco Andreeto, Marc'Aurelio Ranzato, and Pietro Perona. Caltech 101 , Apr 2022

    2022

  14. [22]

    MM-Embed: Universal Multimodal Retrieval with Multimodal LLMs

    Sheng-Chieh Lin, Chankyu Lee, Mohammad Shoeybi, Jimmy Lin, Bryan Catanzaro, and Wei Ping. MM-Embed: Universal Multimodal Retrieval with Multimodal LLMs . arXiv preprint arXiv:2411.02571, 2024

    2024 arXiv

  15. [23]

    LLaVA: Large Language and Vision Assistant

    Haotian Liu, Chunyuan Li, Qingyang Wu, and Yong Jae Lee. LLaVA: Large Language and Vision Assistant . In Advances in Neural Information Processing Systems 36 (NeurIPS 2023), 2023 a

    2023

  16. [24]

    CLIP-driven universal model for organ segmentation and tumor detection

    Jie Liu, Yixiao Zhang, Jie-Neng Chen, Junfei Xiao, Yongyi Lu, Bennett A Landman, Yixuan Yuan, Alan Yuille, Yucheng Tang, and Zongwei Zhou. CLIP-driven universal model for organ segmentation and tumor detection . In Proceedings of the IEEE/CVF international conference on comput...

    2023

  17. [25]

    Universal and Extensible Language-Vision Models for Organ Segmentation and Tumor Detection from Abdominal Computed Tomography

    Jie Liu, Yixiao Zhang, Kang Wang, Mehmet Can Yavuz, Xiaoxi Chen, Yixuan Yuan, Haoliang Li, Yang Yang, Alan Yuille, Yucheng Tang, et al. Universal and Extensible Language-Vision Models for Organ Segmentation and Tumor Detection from Abdominal Computed Tomography . Medical image...

    2024

  18. [26]

    SGDR: Stochastic Gradient Descent with Warm Restarts

    Ilya Loshchilov and Frank Hutter. SGDR: Stochastic Gradient Descent with Warm Restarts . arXiv preprint arXiv:1608.03983, 2016

    2016 arXiv

  19. [27]

    Doubly Right Object Recognition: A Why Prompt for Visual Rationales

    Chengzhi Mao, Revant Teotia, Amrutha Sundar, Sachit Menon, Junfeng Yang, Xin Wang, and Carl Vondrick. Doubly Right Object Recognition: A Why Prompt for Visual Rationales . In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 2672--...

    2023

  20. [28]

    Incremental Learning Techniques for Semantic Segmentation

    Umberto Michieli and Pietro Zanuttigh. Incremental Learning Techniques for Semantic Segmentation . In 2019 IEEE/CVF International Conference on Computer Vision Workshop (ICCVW), pages 3205--3212. IEEE, 2019. doi:10.1109/ICCVW.2019.00400

    2019

  21. [29]

    Scaling Open-Vocabulary Object Detection

    Matthias Minderer, Alexey Gritsenko, and Neil Houlsby. Scaling Open-Vocabulary Object Detection . In Advances in Neural Information Processing Systems, volume 36, pages 72983--73007. Curran Associates, Inc., 2023

    2023

  22. [30]

    CHiLS: Zero-Shot Image Classification with Hierarchical Label Sets

    Zachary Novack, Julian McAuley, Zachary Chase Lipton, and Saurabh Garg. CHiLS: Zero-Shot Image Classification with Hierarchical Label Sets . In International Conference on Machine Learning, pages 26342--26362. PMLR, 2023

    2023

  23. [31]

    RISE: Randomized Input Sampling for Explanation of Black-box Models

    Vitali Petsiuk, Abir Das, and Kate Saenko. RISE: Randomized Input Sampling for Explanation of Black-box Models . arXiv preprint arXiv:1806.07421, 2018

    2018 arXiv

  24. [32]

    Explainable Machine Learning Framework for Image Classification Problems: Case Study on Glioma Cancer Prediction

    Emmanuel Pintelas, Meletis Liaskos, Ioannis E Livieris, Sotiris Kotsiantis, and Panagiotis Pintelas. Explainable Machine Learning Framework for Image Classification Problems: Case Study on Glioma Cancer Prediction . Journal of imaging, 6 0 (6): 0 37, 2020

    2020

  25. [33]

    Object Detection in Self Driving Cars Using Deep Learning

    P Prajwal, D Prajwal, DH Harish, R Gajanana, BS Jayasri, and S Lokesh. Object Detection in Self Driving Cars Using Deep Learning . In 2021 International Conference on Innovative Computing, Intelligent Communication and Smart Electrical Systems (ICSES), pages 1--7. IEEE, 2021

    2021

  26. [34]

    Learning Transferable Visual Models From Natural Language Supervision

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning Transferable Visual Models From Natural Language Supervision . In International conference on machine learning, p...

    2021

  27. [35]

    ECOR: Explainable CLIP for Object Recognition , 2024

    Ali Rasekh, Sepehr Kazemi Ranjbar, Milad Heidari, and Wolfgang Nejdl. ECOR: Explainable CLIP for Object Recognition , 2024. URL https://arxiv.org/abs/2404.12839

    2024 arXiv

  28. [36]

    Why Should I Trust You?

    Marco Tulio Ribeiro, Sameer Singh, and Carlos Guestrin. "Why Should I Trust You?": Explaining the Predictions of Any Classifier . In Proceedings of the 22nd ACM SIGKDD international conference on knowledge discovery and data mining, pages 1135--1144, 2016

    2016

  29. [37]

    Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localization

    Ramprasaath R Selvaraju, Michael Cogswell, Abhishek Das, Ramakrishna Vedantam, Devi Parikh, and Dhruv Batra. Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localization . In Proceedings of the IEEE international conference on computer vision, pages 618--626, 2017

    2017

  30. [38]

    SmoothGrad: removing noise by adding noise

    Daniel Smilkov, Nikhil Thorat, Been Kim, Fernanda Vi \'e gas, and Martin Wattenberg. SmoothGrad: removing noise by adding noise . arXiv preprint arXiv:1706.03825, 2017

    2017 arXiv

  31. [39]

    Axiomatic Attribution for Deep Networks

    Mukund Sundararajan, Ankur Taly, and <bos> Yan. Axiomatic Attribution for Deep Networks . In Proceedings of the 34th International Conference on Machine Learning (ICML), volume 70, pages 3319--3328. PMLR, 2017

    2017

  32. [40]

    C. Wah, S. Branson, P. Welinder, P. Perona, and S. Belongie. The Caltech-UCSD Birds-200-2011 Dataset . Technical Report CNS-TR-2011-001, California Institute of Technology, 2011

    2011

  33. [41]

    SUN Database: Large-scale Scene Recognition from Abbey to Zoo

    Jianxiong Xiao, James Hays, Krista A Ehinger, Aude Oliva, and Antonio Torralba. SUN Database: Large-scale Scene Recognition from Abbey to Zoo . In 2010 IEEE computer society conference on computer vision and pattern recognition, pages 3485--3492. IEEE, 2010

    2010

  34. [42]

    X ^2 -VLM: All-in-One Pre-trained Model For Vision-Language Tasks

    Yan Zeng, Xunjie Chen, Yi Zheng, Bin Wang, Zehuan Liu, Mu Zeng, and Jianwei Zhang. X ^2 -VLM: All-in-One Pre-trained Model For Vision-Language Tasks . arXiv preprint arXiv:2211.12620, 2022

    2022 arXiv

  35. [43]

    Long-CLIP: Unlocking the long-text capability of CLIP

    Beichen Zhang, Pan Zhang, Xiaoyi Dong, Yuhang Zang, and Jiaqi Wang. Long-CLIP: Unlocking the long-text capability of CLIP . In European Conference on Computer Vision, pages 310--325. Springer, 2024 a

    2024

  36. [44]

    Multimodal Chain-of-Thought Reasoning in Language Models

    Zhuosheng Zhang, Xiang Hua, Vilson Singh, Fei Jiang, Hao Chen, Ziqi Lim, and Meng Zhang. Multimodal Chain-of-Thought Reasoning in Language Models . In The Twelfth International Conference on Learning Representations (ICLR 2024), 2024 b

    2024

  37. [45]

    write newline

    " write newline "" before.all 'output.state := FUNCTION fin.entry add.period write newline FUNCTION new.block output.state before.all = 'skip after.block 'output.state := if FUNCTION new.sentence output.state after.block = 'skip output.state before.all = 'skip after.sentence '...

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.