Pith. sign in

REVIEW 4 major objections 4 minor 39 references

Multi-View Incongruity Learning for Multimodal Sarcasm Detection

T0 review · 4 major / 4 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 This paper claims that learning incongruity at token-patch, entity-object, and sentiment levels, fusing the views by estimated credibility, and training with dual text-image augmentation reduces reliance on spurious correlations in…

desk verdict A useful MMSD result and a sensible three-view incongruity design, but the robustness-to-spurious-correlations claim rests on an unvalidated, unreleased LLM-generated test set and does not hold up as reported. read the letter →

arxiv 2412.00756 v2 pith:XQLLW4TA submitted 2024-12-01 cs.CL

Diandian Guo , Cong Cao , Fangfang Yuan , Yanbing Liu , Guangjie Zeng , Xiaoyan Yu , Hao Peng , Philip S. Yu This is my paper · ORCID
classification cs.CL
keywords multimodalsarcasmdetectionspuriouscorrelationsincongruitylearningmulti-viewcontrastivedataaugmentationOCRtextout-of-distributiontestset
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

Existing multimodal sarcasm detectors often get the right answer for the wrong reason, latching onto textual patterns while ignoring images or fixating on non-critical words instead of emotional ones. This paper argues that the remedy is to learn incongruity explicitly in three views—token-patch, entity-object, and sentiment—then fuse the views by estimated credibility, and to train with balanced text and image augmentation plus contrastive learning. To demonstrate generalizability, the authors build SPMSD, a 1,000-sample test set of transformed examples designed to expose spurious correlations. They report that MICL outperforms all compared baselines on the standard MMSD benchmark and records a larger lead on SPMSD, which they take as evidence of reduced spurious-correlation reliance. If the claim is right, it gives a concrete, modular recipe for making multimodal sarcasm detection depend on task-relevant incongruities rather than surface cues.

What carries the argument

The engine is a three-view incongruity learner. Token-patch incongruity uses hybrid cross-attention, letting text and image attend to each other symmetrically instead of always treating text as the query. Entity-object incongruity builds a text graph from dependency-parsed entities and an image graph from object regions, then applies graph attention to expose cross-modal mismatches at the object level. Sentiment incongruity compares sentiment polarity extracted from the source text and from OCR text, feeding those scores through a small network. A credibility module estimates how much to trust each view from pre-softmax evidence, and a contrastive loss over augmented text-image pairs pulls same-label examples together while pushing opposite-label examples apart. Refined, manually proofread OCR text is added as a supplementary input so the model can use text embedded in images.

What would settle it

Have independent annotators verify the labels of all 1,000 SPMSD samples and rerun the comparison on the verified subset; if MICL's accuracy lead over DMSD-CL shrinks to the size of its MMSD lead or disappears, the spurious-correlation story is not supported.

Watch

Extended reading notes

Core claim

The paper's central claim is that a method called MICL (Multimodal Incongruities via Contrastive Learning) can make multimodal sarcasm detection both more accurate and less dependent on spurious cues. The method learns three complementary kinds of text-image incongruity: token-patch mismatches through a hybrid attention mechanism, entity-object mismatches through graph attention over semantic graphs, and sentiment mismatches between the text and OCR text read from the image. It then fuses the three views by estimated credibility rather than simple concatenation, enriches the training data with balanced text and image augmentations, and trains the representations with a contrastive loss. On the standard MMSD dataset the paper reports the highest accuracy among compared models, and on its new SPMSD set it reports a larger accuracy lead, which it interprets as evidence that the model has shifted away from spurious correlations.

Load-bearing premise

The robustness claim rests on SPMSD being a faithful probe of natural spurious correlations, but its transformed samples are generated by language models without reported human validation of the new labels.

Editorial extensions

If this is right

  • MICL reports 92.08% accuracy and 91.81% macro-F1 on MMSD, outperforming all baselines including the strongest compared model, G2SAM, at 91.07% accuracy.
  • On SPMSD, MICL reports 68.7% accuracy versus 60.6% for DMSD-CL, a larger relative advantage than on MMSD, which the paper interprets as stronger generalization.
  • Ablation results show each added incongruity view raises both MMSD and SPMSD accuracy; sentiment incongruity contributes most on SPMSD, while entity-object incongruity contributes most on MMSD.
  • Credibility-weighted fusion improves results on both datasets, and the dual text-image augmentation raises SPMSD accuracy while leaving MMSD accuracy essentially unchanged.
  • Refined OCR-text from images helps all model variants on both datasets, whereas noisier OCR from prior work does not consistently help.

Reading between the lines

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

  • If the robustness gain transfers, the same three-view incongruity recipe could apply to other incongruity-driven multimodal tasks, such as fake news or hate speech detection, where models also lean on surface cues.
  • A decisive isolation test would remove the sentiment view and flip emotional words in the test set; the paper's own ablations suggest sentiment drives much of the SPMSD gain, so this would reveal whether the mechanism is true incongruity modeling or a sentiment shortcut.
  • The SPMSD construction protocol—language-model-based regeneration, text-image swapping, and unimodal variants—could be reused as a cheap robustness audit for any multimodal classifier before building a human-labeled distribution-shift set.
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 / 4 minor

Summary. The paper proposes MICL, a multimodal sarcasm detection method that learns incongruities from three views (token-patch, entity-object, and sentiment), combines them via evidence-based credibility fusion, and augments both text and image training data. The authors also construct a new test set, SPMSD, to probe reliance on spurious correlations. On the public MMSD benchmark, MICL reports the best accuracy (92.08%) among compared methods, and on SPMSD it reports a large accuracy gain over baselines (68.7% vs. 60.6% for DMSD-CL). The central claims are that multi-view incongruity learning plus dual augmentation reduces spurious-correlation reliance and improves generalization.

Significance. If the central claims are substantiated, the paper makes a useful contribution: it draws attention to two concrete failure modes in multimodal sarcasm detection, proposes a multi-view architecture that goes beyond token-patch incongruity, and attempts to construct an evaluation set for spurious correlations. The paper also includes a reasonably thorough ablation study, a comparison with LVLMs, and an analysis of different text encoders. However, the significance is currently limited by verification gaps: the SPMSD set is neither released nor validated by human annotators, the reported gains lack statistical testing, and the "incongruity" modules are not formalized as incongruity computations. These issues are load-bearing for the paper's robustness claim.

major comments (4)
  1. [§3.2.2 and §3.2.3, Eqs. (10) and (12)] The central robustness claim that MICL "can effectively mitigate reliance on spurious correlations, showing better generalization ability on new data" is supported only by SPMSD, a test set constructed by the authors using LLM-driven transformations (sentiment reversal, entity replacement, sarcastic-text regeneration, text-image swapping, and unimodal inputs). Appendix A reports no human validation of the transformed labels, no inter-annotator agreement, and no release of the dataset. Moreover, §3.4 trains MICL on augmented data generated with the same tool families (ChatGPT for text, GLM-4V/Stable Diffusion for images). This creates a confound: MICL's 8.1-point advantage over DMSD-CL on SPMSD may reflect familiarity with generator artifacts or label-distribution shifts rather than genuine robustness to natural spurious correlations. The authors should release SPMSD, report human validation of label preservation, and ideally evaluate on an independently constructed out-of-distribution benchmark before the robustness claim can be accepted.
  2. [Table 1 and §4.3] The modules named "entity-object incongruity learning" and "sentiment incongruity learning" do not compute an explicit incongruity. Eq. (10) is a softmax-weighted sum over the concatenated text and image graph node features, and Eq. (12) is an MLP over sentiment polarities and text hidden states. Neither equation defines a contrast, difference, or mismatch between modalities or between expected and observed sentiment. If incongruity is meant to be captured implicitly through the downstream classifier, the paper should state this and justify why the weighted-sum pooling in Eq. (10) specifically encodes an incongruity rather than a generic multimodal feature. Without a formal definition, the central architectural claim that MICL "learns incongruities from three views" is not actually demonstrated.
  3. [§4.3] The reported MMSD improvement over the reproduced G2SAM* is 1.01 points in accuracy (92.08 vs. 91.07) and 1.03 points in macro-F1. No code, data, random seeds, or significance tests are provided, and the SPMSD results in Table 2 are likewise single-run numbers. Given the small margin over the strongest baseline, the authors should report the mean and standard deviation over multiple runs and perform a significance test (e.g., paired bootstrap or approximate randomization). This is necessary to determine whether the observed differences are stable rather than due to seed or initialization variance.
  4. [§4.5, Table 4] The data augmentation description does not specify how the generated samples are verified or filtered. For example, the text augmentation prompt asks the LLM to reverse or preserve sarcasm, but no human or automatic check is described for whether the generated label matches the intended semantics. Since the contrastive loss in Eq. (15) treats same-label augmented samples as positives and different-label samples as negatives, label errors in augmented data directly corrupt the training signal. The authors should report a validation step for augmented samples or at least provide a random sample of augmented data with human judgments.
minor comments (4)
  1. [§3.2.2] There is a typo in the abstract: "evaluate the the model's generalizability" should be "evaluate the model's generalizability." Similarly, Section 1 contains "to a greater extend" which should be "to a greater extent."
  2. [Figure 5] The text says "If there is a dependency between two entites," where "entites" should be "entities." Also, Eq. (9) writes α_{i,j}W^l g_j^l but the definition of α in Eq. (8) already includes LeakyReLU and an exponential; the notation is understandable but should be consistently indexed across Eq. (8) and Eq. (9).
  3. [Appendix C] The credibility study in Figure 5 is difficult to interpret because the subplot labels appear as unicode fragments (e.g., "/uni00000013/uni00000011/...") rather than readable scenario names. The figure should be regenerated with clear labels for the sarcastic, non-sarcastic, and SPMSD scenarios, and the y-axis should be defined.
  4. [Section 4.4] In the LVLM comparison, the model name "LLaV A 1.6" is likely a typo for "LLaVA 1.6." The paper should also clarify whether the LVLMs receive the same OCR-text as MICL; otherwise the comparison is not apples-to-apples, since MICL uses OCR-text generated by GLM-4V with manual proofreading.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper's central MMSD result is externally benchmarked, and the SPMSD-based robustness claim, while limited by LLM-generated test data, does not reduce to the model's training inputs by construction.

full rationale

The paper's main accuracy claim is established on the public MMSD benchmark (Section 4.3, Table 1), which is external to the authors' construction and not used for parameter fitting. The MICL components (hybrid token-patch attention, GAT-based entity-object incongruity, SenticNet-based sentiment incongruity, and contrastive loss in Eqs. 1-19) are defined from independent inputs and evaluated on held-out MMSD test samples. The robustness conclusion in Section 4.4 rests on SPMSD, a test set constructed by the authors (Appendix A) using LLM-driven transformations: sentiment transformation, entity replacement, text regeneration, text-image swapping, and unimodal inputs. The same transformation families appear in MICL's training augmentation (Section 3.4.1, Appendix I), and no human validation of SPMSD labels is reported. This is a genuine threat to the external validity of the 'better generalization on new data' claim, because SPMSD may sample the augmentation distribution rather than natural spurious correlations. However, it is a benchmark-construction confound rather than a circular derivation: SPMSD is not derived from MICL's outputs or fitted parameters, no SPMSD score enters the training loss, and the MMSD result remains independently informative. The only citation by an overlapping author (Yang et al. 2024) is used for the generic value of multi-view learning and is not load-bearing. No circular step satisfies the evidentiary bar of equation-level reduction or fitted-input renaming.

Assumptions & free parameters 5 free parameters · 5 assumptions · 1 invented entities

The method depends on several hand-set hyperparameters and on unvalidated assumptions about label quality, OCR quality, and the SPMSD benchmark. The largest burden is the SPMSD test set: because it is both the evidence for robustness and constructed by the authors with LLM transformations similar to the training augmentations, it is not an independent yardstick.

free parameters (5)
  • contrastive temperature tau = 0.07
    Set in Appendix F; temperature in Eq. (15) controls the contrastive loss sharpness and is chosen by hand.
  • contrastive loss weight lambda = 1
    Set in Appendix F; balances L_ce and L_cl in Eq. (19).
  • visual graph edge cosine threshold = 0.6
    Appendix F: edges between image regions are added when cosine similarity exceeds 0.6; no sensitivity analysis is reported.
  • text augmentation ratio = 1:1 opposite/same label
    Section 3.4.1: paraphrases and label-flipping generated at 1:1 for all samples; no ablation on the ratio.
  • image augmentation ratio = 3:3:2:2 cropping, swapping, style, generation
    Section 3.4.1: four strategies applied at 3:3:2:2 ratio; chosen by hand.
assumptions (5)
  • domain assumption MMSD labels derived from hashtags are correct sarcasm labels.
    Experiment sections assume hashtag-based labels are ground truth, as in prior MSD work.
  • ad hoc to paper The two failure modes in Figure 1 are the primary causes of spurious correlations in MSD.
    The introduction claims two primary causes based on example perturbations of one SOTA model, without a systematic study.
  • domain assumption SenticNet polarity scores for source text and OCR-text are accurate enough to provide sentiment incongruity.
    Eq. (11) feeds SenticNet outputs into the sentiment view; no validation is provided on this corpus.
  • domain assumption GLM-4V generated OCR-text, after manual proofreading, improves over prior OCR and does not introduce systematic noise.
    Section 3.1.1 claims refined OCR; only qualitative examples in Figure 3 and Appendix H support this.
  • ad hoc to paper The SPMSD construction strategy preserves the original labels while perturbing non-essential features.
    Appendix A lists transformation strategies but no human agreement or label validation, so label preservation is assumed.
invented entities (1)
  • SPMSD test set
    purpose: Evaluate robustness to potential spurious correlations in multimodal sarcasm detection.
    The set is not released, construction details are underspecified, and transformations are generated with LLMs; it has no external validation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Multi-View Incongruity Learning for Multimodal Sarcasm Detection." pith.science (2026). https://pith.science/paper/XQLLW4TA

@misc{pith2026241200756,
  author       = {Pith},
  title        = {Pith review of: Multi-View Incongruity Learning for Multimodal Sarcasm Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XQLLW4TA}},
  note         = {Machine review of arXiv:2412.00756}
}
read the original abstract

Multimodal sarcasm detection (MSD) is essential for various downstream tasks. Existing MSD methods tend to rely on spurious correlations. These methods often mistakenly prioritize non-essential features yet still make correct predictions, demonstrating poor generalizability beyond training environments. Regarding this phenomenon, this paper undertakes several initiatives. Firstly, we identify two primary causes that lead to the reliance of spurious correlations. Secondly, we address these challenges by proposing a novel method that integrate Multimodal Incongruities via Contrastive Learning (MICL) for multimodal sarcasm detection. Specifically, we first leverage incongruity to drive multi-view learning from three views: token-patch, entity-object, and sentiment. Then, we introduce extensive data augmentation to mitigate the biased learning of the textual modality. Additionally, we construct a test set, SPMSD, which consists potential spurious correlations to evaluate the the model's generalizability. Experimental results demonstrate the superiority of MICL on benchmark datasets, along with the analyses showcasing MICL's advancement in mitigating the effect of spurious correlation.

Figures

Figures reproduced from arXiv: 2412.00756 by the authors.

Figure 1
Figure 1. Existing models suffer from two deficiencies [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. The overall architecture of MICL primarily comprises three key modules: (a) Multimodal Feature [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. In the first example, since the text is in [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Summary of text and visual augmentation methods. Text augmentation generates samples with the same or opposite labels. Visual augmentation methods include: (a) cropping, (b) swapping images, (c) image generation, and (d) image style transfer. 3.3 Multi-View Fusion As s…
Figure 5
Figure 5. Figure 5: Credibility study. differences from the original data. For example, MILNet+aug’ slightly improves performance on the SPMSD dataset but causes a decrease on the MMSD dataset. 2) Our OCR-text can enhance the models’ performance. All methods show better results on both be…
Figure 6
Figure 6. Figure 6: Case studies on spuriously correlated samples. [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 7
Figure 7. Figure 7: Attention Visualization. G Prompts Prompts for OCR. Please perform OCR on this image and translate any non-English text into En￾glish. Prompts for Text Augmentation. Please rewrite these data from three aspects: 1. Reverse the mean￾ing of sarcasm: that is, if the sarca…
Figure 9
Figure 9. Figure 9: Data augmentation examples [PITH_FULL_IMAGE:figures/full_fig_p013_9.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

39 extracted references · 27 canonical work pages

  1. [1]

    Peter Anderson, Xiaodong He, Chris Buehler, Damien Teney, Mark Johnson, Stephen Gould, and Lei Zhang. 2018. Bottom-up and top-down attention for image captioning and visual question answering. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 6077--6086

    2018

  2. [2]

    Yitao Cai, Huiyu Cai, and Xiaojun Wan. 2019. Multi-modal sarcasm detection in twitter with hierarchical fusion model. In Proceedings of the 57th annual meeting of the association for computational linguistics, pages 2506--2515

    work page 2019

  3. [3]

    Erik Cambria, Xulang Zhang, Rui Mao, Melvin Chen, and Kenneth Kwok. 2024. Senticnet 8: Fusing emotion ai and commonsense ai for interpretable, trustworthy, and explainable affective computing. In International Conference on Human-Computer Interaction (HCII)

    work page 2024

  4. [4]

    Dmitry Davidov, Oren Tsur, and Ari Rappoport. 2010. Semi-supervised recognition of sarcasm in twitter and amazon. In Proceedings of the fourteenth conference on computational natural language learning, pages 107--116

    work page 2010

  5. [5]

    Yihe Deng, Yu Yang, Baharan Mirzasoleiman, and Quanquan Gu. 2024. Robust learning with progressive data expansion against spurious correlation. Advances in Neural Information Processing Systems, 36

    work page 2024

  6. [6]

    Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. Bert: Pre-training of deep bidirectional transformers for language understanding. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers), pages 4171--4186

    2019

  7. [7]

    Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, et al. 2020. An image is worth 16x16 words: Transformers for image recognition at scale. In International Conference on Learning Representations

    2020

  8. [8]

    DI Hern \'a ndez Farias and Paolo Rosso. 2017. Irony, sarcasm, and sentiment analysis. In Sentiment Analysis in Social Networks, pages 113--128. Elsevier

    work page 2017

Show all 39 references
  1. [9]

    Alex Graves and J \"u rgen Schmidhuber. 2005. Framewise phoneme classification with bidirectional lstm and other neural network architectures. Neural networks, 18(5-6):602--610

    2005

  2. [10]

    Zongbo Han, Changqing Zhang, Huazhu Fu, and Joey Tianyi Zhou. 2021. Trusted multi-view classification. In International Conference on Learning Representations

    2021

  3. [11]

    Mengzhao Jia, Can Xie, and Liqiang Jing. 2024. Debiasing multimodal sarcasm detection with contrastive learning. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 38, pages 18354--18362

    2024

  4. [12]

    Aditya Joshi, Vinita Sharma, and Pushpak Bhattacharyya. 2015. Harnessing context incongruity for sarcasm detection. In Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Process...

    2015

  5. [13]

    Arpit Khare, Amisha Gangwar, Sudhakar Singh, and Shiv Prakash. 2023. Sentiment analysis and sarcasm detection in indian general election tweets. In Research Advances in Intelligent Computing, pages 253--268. CRC Press

    2023

  6. [14]

    Yoon Kim. 2014. Convolutional neural networks for sentence classification. arXiv preprint arXiv:1408.5882

    2014 arXiv

  7. [15]

    Polina Kirichenko, Pavel Izmailov, and Andrew Gordon Wilson. 2023. Last layer re-training is sufficient for robustness to spurious correlations. ICLR 2023

    2023

  8. [16]

    Bin Liang, Chenwei Lou, Xiang Li, Lin Gui, Min Yang, and Ruifeng Xu. 2021. Multi-modal sarcasm detection with interactive in-modal and cross-modal graphs. In Proceedings of the 29th ACM international conference on multimedia, pages 4707--4715

    2021

  9. [17]

    Bin Liang, Chenwei Lou, Xiang Li, Min Yang, Lin Gui, Yulan He, Wenjie Pei, and Ruifeng Xu. 2022. Multi-modal sarcasm detection via cross-modal graph convolutional network. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Lon...

    2022

  10. [18]

    Yinhan Liu, Myle Ott, Naman Goyal, Jingfei Du, Mandar Joshi, Danqi Chen, Omer Levy, Mike Lewis, Luke Zettlemoyer, and Veselin Stoyanov. 2019. Roberta: A robustly optimized bert pretraining approach. arXiv preprint arXiv:1907.11692

    2019 arXiv

  11. [19]

    Zihan Ma, Minnan Luo, Hao Guo, Zhi Zeng, Yiran Hao, and Xiang Zhao. 2024. Event-radar: Event-driven multi-view learning for multimodal fake news detection. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages...

    2024

  12. [20]

    Yulei Niu, Kaihua Tang, Hanwang Zhang, Zhiwu Lu, Xian-Sheng Hua, and Ji-Rong Wen. 2021. Counterfactual vqa: A cause-effect look at language bias. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 12700--12710

    2021

  13. [21]

    Hongliang Pan, Zheng Lin, Peng Fu, Yatao Qi, and Weiping Wang. 2020. Modeling intra and inter-modality incongruity for multi-modal sarcasm detection. In Findings of the Association for Computational Linguistics: EMNLP 2020, pages 1383--1392

    2020

  14. [22]

    MSM Prasanna, SG Shaila, and A Vadivel. 2023. Polarity classification on twitter data for classifying sarcasm using clause pattern for sentiment analysis. Multimedia Tools and Applications, 82(21):32789--32825

    2023

  15. [23]

    Yang Qiao, Liqiang Jing, Xuemeng Song, Xiaolin Chen, Lei Zhu, and Liqiang Nie. 2023. Mutual-enhanced incongruity learning network for multi-modal sarcasm detection. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 37, pages 9507--9515

    2023

  16. [24]

    Libo Qin, Shijue Huang, Qiguang Chen, Chenran Cai, Yudi Zhang, Bin Liang, Wanxiang Che, and Ruifeng Xu. 2023. Mmsd2. 0: Towards a reliable multi-modal sarcasm detection system. In Findings of the Association for Computational Linguistics: ACL 2023, pages 10834--10845

    2023

  17. [25]

    Rossano Schifanella, Paloma De Juan, Joel Tetreault, and Liangliang Cao. 2016. Detecting sarcasm in multimodal social platforms. In Proceedings of the 24th ACM international conference on Multimedia, pages 1136--1145

    2016

  18. [26]

    P Veli c kovi \'c , A Casanova, P Li \`o , G Cucurull, A Romero, and Y Bengio. 2018. Graph attention networks

    2018

  19. [27]

    Weihan Wang, Qingsong Lv, Wenmeng Yu, Wenyi Hong, Ji Qi, Yan Wang, Junhui Ji, Zhuoyi Yang, Lei Zhao, Xixuan Song, Jiazheng Xu, Bin Xu, Juanzi Li, Yuxiao Dong, Ming Ding, and Jie Tang. 2023. https://arxiv.org/abs/2311.03079 Cogvlm: Visual expert for pretrained language models ....

    2023 arXiv

  20. [28]

    Yifei Wang, Jizhe Zhang, and Yisen Wang. 2024. Do generated data always help contrastive learning? arXiv preprint arXiv:2403.12448

    2024 arXiv

  21. [29]

    Yiwei Wei, Shaozu Yuan, Hengyang Zhou, Longbiao Wang, Zhiling Yan, Ruosong Yang, and Meng Chen. 2024. G\^ 2sam: Graph-based global semantic awareness method for multimodal sarcasm detection. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 38, pages 9151--9159

    2024

  22. [30]

    Changsong Wen, Guoli Jia, and Jufeng Yang. 2023. Dip: Dual incongruity perceiving network for sarcasm detection. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 2540--2550

    2023

  23. [31]

    Zixin Wen and Yuanzhi Li. 2021. Toward understanding the feature learning process of self-supervised contrastive learning. In International Conference on Machine Learning, pages 11112--11122. PMLR

    2021

  24. [32]

    Nan Wu, Stanislaw Jastrzebski, Kyunghyun Cho, and Krzysztof J Geras. 2022. Characterizing and overcoming the greedy nature of learning in multi-modal deep neural networks. In International Conference on Machine Learning, pages 24043--24055. PMLR

    2022

  25. [33]

    Tao Xiong, Peiran Zhang, Hongbo Zhu, and Yihui Yang. 2019. Sarcasm detection with self-matching networks and low-rank bilinear pooling. In The world wide web conference, pages 2115--2124

    2019

  26. [34]

    Nan Xu, Zhixiong Zeng, and Wenji Mao. 2020. Reasoning with multimodal sarcastic tweets via modeling cross-modality contrast and semantic association. In Proceedings of the 58th annual meeting of the association for computational linguistics, pages 3777--3786

    2020

  27. [35]

    Yingguang Yang, Qi Wu, Buyun He, Hao Peng, Renyu Yang, Zhifeng Hao, and Yong Liao. 2024. Sebot: Structural entropy guided multi-view contrastive learning for social bot detection. In Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, pages 3841--3852

    2024

  28. [36]

    Sriram Yenamandra, Pratik Ramesh, Viraj Prabhu, and Judy Hoffman. 2023. Facts: First amplify correlations and then slice to discover bias. In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), pages 4794--4804

    2023

  29. [37]

    Meishan Zhang, Yue Zhang, and Guohong Fu. 2016. Tweet sarcasm detection using deep neural network. In Proceedings of COLING 2016, the 26th International Conference on Computational Linguistics: technical papers, pages 2449--2460

    2016

  30. [38]

    online" 'onlinestring :=

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint eprinttype howpublished institution journal key month note number organization pages publisher school series title type volume year doi pubmed url lastchecked label extra.label sort.label short.list...

  31. [39]

    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...

Pith tools

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