Pith. sign in

REVIEW 5 major objections 5 minor 27 references

Hierarchical Text Classification Using Contrastive Learning Informed Path Guided Hierarchy

T0 review · 5 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read A contrastive text encoder plus a path-guided hierarchy classifier beats either alone.

desk verdict A genuine but incremental combination of two HTC approaches; the abstract overstates the SOTA gains because the upper range rests on re-implemented baselines. read the letter →

arxiv 2506.04381 v1 pith:ZTFBR4QP submitted 2025-06-04 cs.CL cs.LG

Neeraj Agrawal , Saurabh Kumar , Priyanka Bhatt , Tanishka Agarwal This is my paper
classification cs.CLcs.LG
keywords hierarchicaltextclassificationcontrastivelearningpath-guidedhierarchymulti-labelBERTGraphormermaximumpoolingpositivesamplegeneration
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

Hierarchical text classification assigns documents to nodes in a tree or DAG of categories, and existing systems generally take one of two routes: they either learn a separate encoder for the label hierarchy and mix it with text features, or they infuse hierarchy directly into the text encoder with contrastive learning. This paper claims those two routes capture complementary information, and proposes HTC-CLIP, an architecture that trains both at once: a linear classifier and a path-guided hierarchy classifier on a contrastively trained BERT encoder. At inference the model max-pools the two probability distributions per class. On the WOS and NYT benchmarks it reports Macro-F1 of 81.64 and 68.59, with improvements of roughly 0.99 to 2.37 percent over the state-of-the-art baselines it compares against. If correct, the field no longer has to treat the two hierarchy-encoding strategies as alternatives; they can be combined for small but consistent gains.

What carries the argument

The load-bearing objects are two classifiers trained on one BERT encoder. The path-guided hierarchy classifier predicts each level from the [CLS] vector and from the previous level's activations through linear-ReLU hidden layers, then sums the two activation streams for each level and concatenates across levels before sigmoid; this is the explicit structural head. The linear classifier is a single projection to all classes. Both heads are trained on original text and on positive samples obtained by keeping tokens with above-threshold Gumbel-Softmax label-attention probabilities, with an NT-Xent contrastive loss aligning original and positive [CLS] representations. Graphormer encodes the label hierarchy into the BERT text encoder. At inference, per-class max pooling combines the two heads' probabilities. The chained activations carry parent-to-child information; the contrastive objective carries label-hierarchy information into the text encoding.

What would settle it

Run HGCLR's original implementation or checkpoints on the same WOS and NYT splits with the same BERT base and compare Micro-F1 and Macro-F1 to the numbers in Table 2; if faithful HGCLR reaches or exceeds 87.86/81.64 on WOS or 79.22/68.59 on NYT, the claimed improvement over the state of the art does not hold.

Watch

Extended reading notes

Core claim

The paper's central claim is that hierarchy-aware text encoding and explicit path-guided hierarchy classification are complements, not substitutes. HTC-CLIP learns a BERT representation whose [CLS] output is trained with contrastive loss against positive samples formed by Gumbel-Softmax token dropping guided by label attention; this same representation feeds a linear classifier and a chained per-level hierarchical classifier. Training minimizes binary cross-entropy on both heads for both original and positive samples, plus a contrastive term. At inference, the two sigmoid probability vectors are max-pooled per class. The reported result is Micro-F1/Macro-F1 of 87.86/81.64 on WOS and 79.22/68.59 on NYT, beating the reimplemented HGCLR baseline by 0.85/1.34 and 0.88/2.37 points respectively. The ablations show that removing either head, removing the hidden layers in the path-guided head, or removing the positive-sample BCE losses degrades validation F1.

Load-bearing premise

The reported gains depend on the authors' BERT and HGCLR re-implementations being faithful baselines, since the paper states it could not reproduce the numbers in [20]; a weaker baseline would inflate the improvement over prior work.

Editorial extensions

If this is right

  • The two previously competing families of HTC methods, mixed-representation models and hierarchy-infused text encoders, can be combined in one end-to-end architecture rather than chosen between.
  • Max pooling of the two heads outperforms either head alone and also average pooling, so per-label disagreement between the heads appears to carry signal.
  • The positive-sample BCE losses are load-bearing: removing them drops validation Macro-F1 by roughly 0.7 to 1.7 points, indicating contrastively learned positives help tune both classifiers.
  • The parameter cost of adding the path-guided hierarchy head is small, about 0.2 million parameters over the contrastive encoder, so the gains do not require a much larger model.
  • The gains appear on both a shallow hierarchy (WOS, depth 2) and a deep one (NYT, depth 8), suggesting the mechanism is not tied to one tree depth.

Reading between the lines

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

  • If the original HGCLR results are reproducible, the reported margin over published state of the art may narrow, but the internal comparison between the two heads would still show complementarity within this implementation.
  • The architecture suggests a general recipe: train one encoder with implicit structure via contrastive loss and one explicit structured head, then ensemble by max pooling, which could transfer to other hierarchical prediction tasks such as taxonomy-aware product categorization or medical coding.
  • A testable extension is to analyze cases where the two heads disagree most; if the path-guided head wins on deep or tail labels and the linear head wins on shallow frequent labels, max pooling could be replaced by a learned gating function.
  • The Gumbel-Softmax token-dropping positive sampler could be replaced by label-driven span masking, and if performance holds, token-level attribution would not be the only viable positive construction for the contrastive objective.
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

5 major / 5 minor

Summary. The paper proposes HTC-CLIP, a hierarchical text classification model that combines two existing ideas: contrastive learning that encodes label hierarchy into a BERT-based text encoder (following HGCLR) and an explicit path-guided hierarchy classifier built from chained linear layers over the pooled text representation. Two classifiers are trained simultaneously, and at inference their sigmoid probability outputs are combined by element-wise maximum pooling. Experiments on WOS and NYT report Micro-F1 and Macro-F1 scores, and the abstract claims a 0.99–2.37% Macro-F1 improvement over state-of-the-art models. The central empirical claim is that the two classifier heads capture complementary information and that pooling their outputs yields better hierarchical text classification than either head alone.

Significance. If the empirical claims hold, HTC-CLIP is a useful, low-overhead combination of two previously separate approaches: the paper's Table 3 shows only a negligible parameter increase over HGCLR, and the ablations in Tables 4–6 provide some evidence that both the contrastive-learning-informed classifiers and the path-guided hierarchy contribute. The idea of training two complementary heads and max-pooling their probabilities at inference is simple and reproducible in principle. However, the headline state-of-the-art claim is not currently supported as stated: the abstract's improvement range derives from comparisons with the authors' own re-implementations of BERT and HGCLR rather than with published baseline numbers, and the statistical evidence for significance is incomplete. The contribution is therefore plausible but needs stronger baseline verification and more rigorous reporting before it can be accepted as a demonstrated improvement over existing methods.

major comments (5)
  1. [Abstract and Section 5.2, Table 2] The abstract's claim of '0.99 - 2.37% in Macro F1' over state-of-the-art models is not supported by Table 2 as presented. On WOS, the best external baseline shown, BERT+HiMatch, has Macro-F1 81.06 versus HTC-CLIP's 81.64, a difference of 0.58 percentage points; on NYT, BERT+HTCInfoMax has 67.31 versus 68.59, a difference of 1.28 percentage points. The upper end of the claimed range, 2.37, is exactly the Macro-F1 gap between HTC-CLIP and the authors' re-implemented HGCLR on NYT (68.59 vs. 66.22). Since Section 5.2 states 'We could not reproduce the BERT and HGCLR results reported in [20]', the comparative evidence against the actual published state of the art is missing. The authors should report the published baseline numbers alongside their re-implementations, or provide evidence that their re-implementations match the published models' performance.
  2. [Section 5.2, statistical significance] The t-test claim is not adequately substantiated. The text reports that the WOS experiment was run five times and gives standard deviations of 0.26 for Micro-F1 and 0.17 for Macro-F1, but it does not state whether these standard deviations are for HTC-CLIP, HGCLR, or both, and it gives no p-values or test statistic. A t-test comparing a five-run sample to a single baseline run would not be valid, and the claim that the means are 'more than three standard deviations away' conflates standard deviation of one model with the standard error of the difference. NYT has no repeated runs or variance reported at all, yet the text says the model 'performs significantly better' on that dataset. The authors should report the full experimental protocol, including number of runs for each model, and appropriate significance tests with effect sizes for both datasets.
  3. [Section 4.2.2, Eq. (13)] The positive sample construction uses a non-differentiable thresholding operation, but the paper only says 'we implement it differently to make sure the whole model can be trained end-to-end' without specifying the actual mechanism. Since the contrastive loss and the positive-sample classification losses are central to the method, the gradient flow through the masking operation must be described concretely (e.g., straight-through estimator, soft masking, or stop-gradient). Without this detail, the training procedure is not fully reproducible.
  4. [Section 4.2.1, Eq. (8)] The notation in Eq. (8) is inconsistent with the surrounding text. The text states that 'all activations from each level are concatenated and passed through sigmoidal activation', but Eq. (8) writes P_c = sigma(A_1 ⊙ A_2 ⊙ ... ⊙ A_|H|). The activations A_h have dimension C_h, the number of classes at level h, which differ across levels; a Hadamard product is not well-defined across different dimensions. If the intended operation is concatenation after some flattening or projection, that should be stated explicitly, because the path-guided hierarchy classifier is a load-bearing component of the architecture.
  5. [Section 5.3.3, Table 6 and Section 4.4] The pooling procedure is described ambiguously. Section 4.4 says that during testing 'the maximum pool output of both the linear classifier and path-guided classifier is used', but Table 6 contains two rows both labeled 'Both l.c. & h.c. with max pool', with different scores (87.01/80.42 and 87.58/80.88), and the text says that adding or taking the maximum during training 'is not able to outperform' the proposed architecture where max pooling is done during inference. The authors should clarify exactly which pooling is applied during training and which during inference, and why the two rows in Table 6 differ.
minor comments (5)
  1. [Table 1 and Section 5.1] Table 1's caption says 'Statistics of three datasets' and Section 5.1 says the hidden layer size k is 'set to 128 for all three datasets', but only WOS and NYT are used or listed. This appears to be an editorial remnant and should be corrected.
  2. [References] References [2] and [3] are identical in title, authors, venue, and pages. Since the text cites [3] as HiMatch and [2] as a separate hierarchy-aware label semantics matching network, one of these citations is almost certainly wrong and should be corrected.
  3. [Section 4.2.1, Eq. (7)] Equation (7) uses the symbol ⊕ while the preceding sentence says the final activation is the 'sum' of two activations. If element-wise addition is intended, the symbol should be +; if concatenation is intended, the dimensions in later equations do not work. Please clarify.
  4. [Section 4.2.2, Eq. (10)] The label features l_j used in Eq. (10) are never defined. The paper should specify how the label embedding matrix is initialized and whether it is learned jointly with the model.
  5. [Table 2] The NYT row for BERT+HiMatch is missing. The paper should either provide the result or explain why HiMatch is not evaluated on NYT.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: HTC-CLIP is an empirical architectural combination validated on held-out test data, with no load-bearing self-citation or definitional reduction.

full rationale

The paper's derivation chain is architectural and empirical, not definitional. The method combines a contrastive-learning-guided BERT text encoder (adopted from external prior work, HGCLR) with a separate path-guided hierarchy classifier, trains both with BCE losses and a contrastive loss (Eq. 27), and pools the two classifiers' probability outputs via max pooling at inference. Each component is specified by explicit equations (Eqs. 3-8, 15-19, 20-26), and the central claim — that the two heads capture complementary information and that pooling improves results — is supported by held-out test performance in Table 2 and by ablations in Tables 4-6. No parameter is fitted to the test set, and the reported hyperparameters (gamma, lambda, hidden size) are validation-tuned and disclosed. The comparison to HGCLR and BERT relies on the authors' re-implementations, and the paper explicitly states it could not reproduce the published numbers ('We could not reproduce the BERT and HGCLR results reported in [20], so we report the results of our implementation'). This is a reproducibility/baseline-fidelity concern, not circularity: the re-implemented baselines are not constructed from or equivalent to the proposed model's outputs, and the claimed improvement is not forced by equation or by self-citation. The paper cites [20] for components it reuses, but those are independent external works with no author overlap, so the citation is real evidence rather than a self-supporting chain. There are no uniqueness claims, no renamings of known results, and no fitted quantities renamed as predictions. The only notable weakness is external comparability due to unreproduced baselines, which falls under correctness risk, not circularity.

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

All building blocks are external: BERT for text, Graphormer for hierarchy, NT-Xent for contrastive loss. No new entity is introduced. The paper contributes a wiring and training scheme; its load-bearing choices are summarized by the four free hyperparameters and the domain assumptions listed above.

free parameters (4)
  • lambda, contrastive loss weight = 0.05 WOS, 0.3 NYT
    Weights the contrastive term in Eq. 27 and is tuned per dataset on validation.
  • gamma, positive sample token threshold = 0.02 WOS, 0.005 NYT
    Controls which tokens are retained in positive samples in Eq. 13; tuned per dataset.
  • k, path-guided hidden layer size = 128
    Width of intermediate layers in Eqs. 5 and 6; chosen by hand.
  • batch size = 56
    Training batch size; no sensitivity analysis is reported.
assumptions (5)
  • domain assumption BERT pretrained representations transfer to hierarchical text classification.
    Adopted in Section 4.1 without re-validation.
  • domain assumption Label hierarchies can be treated as trees or DAGs and all levels can be flattened into |C| binary outputs.
    Assumed in Section 3 and Section 4.2.1.
  • domain assumption Contrastive loss on label-attention positive samples improves hierarchy-aware text encoding.
    Taken from HGCLR [20] and used unchanged in Section 4.2.3.
  • domain assumption Graphormer is a suitable hierarchy encoder within the text encoder.
    Borrowed from [20]; no alternative hierarchy encoders are tested.
  • ad hoc to paper The weighted loss sum in Eq. 27 is a valid training objective.
    No theoretical justification is given; the weight lambda is tuned per dataset.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Hierarchical Text Classification Using Contrastive Learning Informed Path Guided Hierarchy." pith.science (2026). https://pith.science/paper/ZTFBR4QP

@misc{pith2026250604381,
  author       = {Pith},
  title        = {Pith review of: Hierarchical Text Classification Using Contrastive Learning Informed Path Guided Hierarchy},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZTFBR4QP}},
  note         = {Machine review of arXiv:2506.04381}
}
read the original abstract

Hierarchical Text Classification (HTC) has recently gained traction given the ability to handle complex label hierarchy. This has found applications in domains like E- commerce, customer care and medicine industry among other real-world applications. Existing HTC models either encode label hierarchy separately and mix it with text encoding or guide the label hierarchy structure in the text encoder. Both approaches capture different characteristics of label hierarchy and are complementary to each other. In this paper, we propose a Hierarchical Text Classification using Contrastive Learning Informed Path guided hierarchy (HTC-CLIP), which learns hierarchy-aware text representation and text informed path guided hierarchy representation using contrastive learning. During the training of HTC-CLIP, we learn two different sets of class probabilities distributions and during inference, we use the pooled output of both probabilities for each class to get the best of both representations. Our results show that the two previous approaches can be effectively combined into one architecture to achieve improved performance. Tests on two public benchmark datasets showed an improvement of 0.99 - 2.37% in Macro F1 score using HTC-CLIP over the existing state-of-the-art models.

Figures

Figures reproduced from arXiv: 2506.04381 by the authors.

Figure 1
Figure 1. Input query is tagged with "missing items" from two different paths "order status" or "refund an item." that optimizes global and local loss. [14] decomposes the hierarchy into subgraphs and conducts Text-GCN on n-gram tokens, whereas [17] applies CNN to utilize the data in the upper levels to contribute to categorization in the lower levels. In early global approaches, HTC is solved by reducing the prob￾lem to a fl… view at source ↗
Figure 2
Figure 2. Different ways of introducing hierarchy information. (a) Previous work of modelling path-guided hierarchy, on top of text encoder [22]. (b) Previous work of modelling text and labels separately and finding a mixed representation [28, 5]. (c) Previous work of incorporating hierarchy information into text encoder for a hierarchy-aware text representation [20]. (d) Our work where the model learns two classifiers, one w… view at source ↗
Figure 3
Figure 3. HTC-CLIP architecture. It contains a BERT based encoder, which generates a hierarchy-aware text representation and a positive sequence output representation. Both representations are passed through two classifiers (linear classifier and path-guided hierarchy classifier) to minimize binary cross-entropy loss, which helps in generating text-aware hierarchy representations. 4.2.3 Contrastive Learning for Text Encoder C… view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

27 extracted references · 26 canonical work pages

  1. [3]

    4370– 4379, Online, (August 2021)

    Haibin Chen, Qianli Ma, Zhenxi Lin, and Jiangyue Yan, ‘Hierarchy- aware label semantics matching network for hierarchical text classifica- tion’, inProceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers), pp. 4370– 4379, Onlin...

    work page 2021

  2. [20]

    7109–7119, Dublin, Ireland, (May 2022)

    Zihan Wang, Peiyi Wang, Lianzhe Huang, Xin Sun, and Houfeng Wang, ‘Incorporating hierarchy into text encoder: a contrastive learn- ing approach for hierarchical text classification’, inProceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pp. 7109–7119, Dublin, Ireland, (May 2022). Association for...

    work page 2022

  3. [1]

    6295–6300, Florence, Italy, (July 2019)

    Siddhartha Banerjee, Cem Akkaya, Francisco Perez-Sorrosal, and Kostas Tsioutsiouliklis, ‘Hierarchical transfer learning for multi-label text classification’, inProceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pp. 6295–6300, Florence, Italy, (July 2019). Association for Computational Linguistics

    work page 2019

  4. [4]

    1597–1607

    Ting Chen, Simon Kornblith, Mohammad Norouzi, and Geoffrey Hin- ton, ‘A simple framework for contrastive learning of visual representa- tions’, inProceedings of the 37th International Conference on Machine Learning, eds., Hal Daumé III and Aarti Singh, volume 119 ofProceed- ings of Machine Learning Research, pp. 1597–1607. PMLR, (13–18 Jul 2020)

    work page 2020

  5. [5]

    3259–3265, Online, (June 2021)

    Zhongfen Deng, Hao Peng, Dongxiao He, Jianxin Li, and Philip Yu, ‘HTCInfoMax: A global model for hierarchical text classification via information maximization’, inProceedings of the 2021 Conference of the North American Chapter of the Association for Computational Lin- guistics: Human Language Technologies, pp. 3259–3265, Online, (June 2021). Association ...

    work page 2021

  6. [6]

    4171–4186, Minneapolis, Minnesota, (June 2019)

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

    work page 2019

  7. [7]

    257–265, New York, NY , USA, (2013)

    Siddharth Gopal and Yiming Yang, ‘Recursive regularization for large- scale classification with hierarchical and graphical dependencies’, in Proceedings of the 19th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, KDD ’13, p. 257–265, New York, NY , USA, (2013). Association for Computing Machinery

    work page 2013

  8. [8]

    Eric Jang, Shixiang Gu, and Ben Poole, ‘Categorical Reparame- terization with Gumbel-Softmax’,arXiv e-prints, arXiv:1611.01144, (November 2016)

    arXiv 2016

Show all 27 references
  1. [9]

    103– 112, Denver, Colorado, (May–June 2015)

    Rie Johnson and Tong Zhang, ‘Effective use of word order for text cat- egorization with convolutional neural networks’, inProceedings of the 2015 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pp. 103– 11...

    2015

  2. [10]

    Brown, Mojtaba Heidarysafa, Kiana Ja- fari Meimandi, Matthew S

    Kamran Kowsari, Donald E. Brown, Mojtaba Heidarysafa, Kiana Ja- fari Meimandi, Matthew S. Gerber, and Laura E. Barnes, ‘Hdltex: Hier- archical deep learning for text classification’, in2017 16th IEEE Inter- national Conference on Machine Learning and Applications (ICMLA), pp. ...

    2017

  3. [11]

    2267–2273

    Siwei Lai, Liheng Xu, Kang Liu, and Jun Zhao, ‘Recurrent convolu- tional neural networks for text classification’, inProceedings of the Twenty-Ninth AAAI Conference on Artificial Intelligence, AAAI’15, p. 2267–2273. AAAI Press, (2015)

    2015

  4. [12]

    445–455, Hong Kong, China, (November 2019)

    Yuning Mao, Jingjing Tian, Jiawei Han, and Xiang Ren, ‘Hierarchi- cal text classification with reinforced label assignment’, inProceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Lan- guag...

    2019

  5. [13]

    Yu Meng, Jiaming Shen, Chao Zhang, and Jiawei Han, ‘Weakly- supervised hierarchical text classification’, inAAAI, (2019)

    2019

  6. [14]

    1063–1072, Republic and Canton of Geneva, CHE, (2018)

    Hao Peng, Jianxin Li, Yu He, Yaopeng Liu, Mengjiao Bao, Lihong Wang, Yangqiu Song, and Qiang Yang, ‘Large-scale hierarchical text classification with recursively regularized deep graph-cnn’, inPro- ceedings of the 2018 World Wide Web Conference, WWW ’18, p. 1063–1072, Republic...

    2018

  7. [15]

    Hao Peng, Jianxin Li, Senzhang Wang, Lihong Wang, Qiran Gong, Renyu Yang, Bo Li, S Yu Philip, and Lifang He, ‘Hierarchical taxonomy-aware and attentional graph capsule rcnns for large-scale multi-label text classification’,IEEE Transactions on Knowledge and Data Engineering,33...

    2019

  8. [16]

    The New York Times Annotated Corpus - Linguistic Data Consortium, 2008

    Evan Sandhaus. The New York Times Annotated Corpus - Linguistic Data Consortium, 2008. [Online; accessed 0000-00-00]

    2008

  9. [17]

    811–816, Brussels, Belgium, (October- November 2018)

    Kazuya Shimura, Jiyi Li, and Fumiyo Fukumoto, ‘HFT-CNN: Learning hierarchical category structure for multi-label short text categorization’, inProceedings of the 2018 Conference on Empirical Methods in Nat- ural Language Processing, pp. 811–816, Brussels, Belgium, (October- No...

    2018

  10. [18]

    Carlos N Silla and Alex A Freitas, ‘A survey of hierarchical classifica- tion across different application domains’,Data Mining and Knowledge Discovery,22(1), 31–72, (2011)

    2011

  11. [19]

    Yangqiu Song and Dan Roth, ‘On dataless hierarchical text classifica- tion’, inProceedings of the AAAI Conference on Artificial Intelligence, volume 28, (2014)

    2014

  12. [21]

    Barros, Silvia N

    Jônatas Wehrmann, Rodrigo C. Barros, Silvia N. das Dôres, and Ri- cardo Cerri, ‘Hierarchical multi-label classification with chained neu- ral networks’, inProceedings of the Symposium on Applied Computing, SAC ’17, p. 790–795, New York, NY , USA, (2017). Association for Comput...

    2017

  13. [22]

    5075–5084

    Jonatas Wehrmann, Ricardo Cerri, and Rodrigo Barros, ‘Hierarchi- cal multi-label classification networks’, inInternational Conference on Machine Learning, pp. 5075–5084. PMLR, (2018)

    2018

  14. [23]

    38–45, Online, (October 2020)

    Thomas Wolf, Lysandre Debut, Victor Sanh, Julien Chaumond, Clement Delangue, Anthony Moi, Pierric Cistac, Tim Rault, Remi Louf, Morgan Funtowicz, Joe Davison, Sam Shleifer, Patrick von Platen, Clara Ma, Yacine Jernite, Julien Plu, Canwen Xu, Teven Le Scao, Sylvain Gugger, Mari...

    2020

  15. [24]

    4354–4364, Hong Kong, China, (November 2019)

    Jiawei Wu, Wenhan Xiong, and William Yang Wang, ‘Learning to learn and predict: A meta-learning approach for multi-label classification’, inProceedings of the 2019 Conference on Empirical Methods in Nat- ural Language Processing and the 9th International Joint Conference on Na...

    2019

  16. [25]

    3915–3926, Santa Fe, New Mexico, USA, (August 2018)

    Pengcheng Yang, Xu Sun, Wei Li, Shuming Ma, Wei Wu, and Houfeng Wang, ‘SGM: Sequence generation model for multi-label classifica- tion’, inProceedings of the 27th International Conference on Com- putational Linguistics, pp. 3915–3926, Santa Fe, New Mexico, USA, (August 2018). ...

    2018

  17. [26]

    Ranzato, A

    Chengxuan Ying, Tianle Cai, Shengjie Luo, Shuxin Zheng, Guolin Ke, Di He, Yanming Shen, and Tie-Yan Liu, ‘Do transformers re- ally perform badly for graph representation?’, inAdvances in Neu- ral Information Processing Systems, eds., M. Ranzato, A. Beygelz- imer, Y . Dauphin, ...

    2021

  18. [27]

    Sheng, Ahoud Alhazmi, and Chenliang Li, ‘Adversarial attacks on deep-learning models in natural language processing: A survey’,ACM Trans

    Wei Emma Zhang, Quan Z. Sheng, Ahoud Alhazmi, and Chenliang Li, ‘Adversarial attacks on deep-learning models in natural language processing: A survey’,ACM Trans. Intell. Syst. Technol.,11(3), (apr 2020)

    2020

  19. [28]

    1106–1117, Online, (July 2020)

    Jie Zhou, Chunping Ma, Dingkun Long, Guangwei Xu, Ning Ding, Haoyu Zhang, Pengjun Xie, and Gongshen Liu, ‘Hierarchy-aware global model for hierarchical text classification’, inProceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pp. 1106–111...

    2020

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.