Pith. sign in

REVIEW 5 major objections 5 minor 42 references

Boosting Neural Language Inference via Cascaded Interactive Reasoning

T0 review · 5 major / 5 minor · reviewed 2026-08-15 · 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 comparing every token pair at every transformer layer, via a stacked interaction tensor, improves natural language inference accuracy across ten benchmarks.

desk verdict The CIRN architecture is a plausible new assembly, but the evaluation is internally inconsistent — the RTE outlier, STS-B metric, and missing robustness experiments break the paper's central claim. read the letter →

arxiv 2505.06607 v1 pith:IR466JZ2 submitted 2025-05-10 cs.CL

Min Li , Chun Yuan This is my paper · ORCID
classification cs.CL
keywords naturallanguageinferencetextualentailmentpre-trainedmodelsmulti-layerrepresentationsinteractiontensorDenseNetfeatureextractionsentence-pairclassificationGLUEbenchmark
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

The paper argues that natural language inference misses useful signal when a model reads only the final layer of a pre-trained transformer. It proposes CIRN, which computes the element-wise product of token representations between the premise and hypothesis at every layer, stacks those interaction tensors, and runs a DenseNet over the stack before classifying. On ten sentence-pair benchmarks, CIRN claims average accuracy gains of 0.8% over BERT-base, 0.7% over BERT-large, 1.5% over RoBERTa-base, and 0.5% over RoBERTa-large. The appeal is that a relatively simple add-on could strengthen any transformer-based pair classifier without modifying the pre-trained backbone.

What carries the argument

The load-bearing object is the stacked interaction tensor: for each layer $l$, every token of the first sentence is multiplied element-wise with every token of the second sentence to form $I^{(l)} \in \mathbb{R}^{n \times m \times d}$, and the $L$ tensors are concatenated along a new layer axis. DenseNet, borrowed from image processing as a feature extractor with dense blocks computing $z_k = f_k(\text{Concat}(Z_0, z_1, \dots, z_{k-1}))$, then compresses this high-order tensor, and a single linear-softmax layer predicts the relation. This machinery replaces terminal-layer alignment with an all-depth comparison, letting the model see fine-grained local differences and global context at the same time.

What would settle it

Run CIRN under the official GLUE evaluation harness, scoring STS-B by Pearson/Spearman correlation and RTE on the standard test set, and re-check whether RoBERTa-base-CIRN actually reaches 82.5 on RTE against a 73.6 baseline. If those numbers do not reproduce, the paper's consistent-improvement claim is falsified.

Watch

Extended reading notes

Core claim

CIRN's central claim is that multi-layer interactive features contain information that terminal-layer representations miss. For each Transformer layer $l$, the model forms the interaction tensor $I^{(l)}_{i,j} = h^{(l,i)}_1 \odot h^{(l,j)}_2$ for every token pair between the two sentences, stacks the $L$ tensors into $I_{\text{stack}} \in \mathbb{R}^{n \times m \times d \times L}$, and compresses them with DenseNet before a softmax classifier. In the paper's Table 1, this moves average accuracy from 85.8 to 86.8 over BERT-base, from 88.0 to 88.8 over BERT-large, from 87.6 to 89.2 over RoBERTa-base, and from 90.3 to 90.7 over RoBERTa-large across the ten datasets, with the largest single swing on RTE for RoBERTa-base (73.6 to 82.5).

Load-bearing premise

The load-bearing premise is that the ten benchmark results in Table 1 were computed under one consistent evaluation protocol, with STS-B scored correctly and the RTE numbers coming from the same setup; if anything was measured differently, the average-gain claim collapses.

Editorial extensions

If this is right

  • If the reported gains are real, adding CIRN to a frozen pre-trained encoder is a plug-in upgrade worth roughly one point of accuracy on sentence-pair benchmarks.
  • The same module transfers to paraphrase identification when paraphrase pairs are reframed as entailment, so the method is claimed to be task-general.
  • The ablation results attribute about 1.2 points of accuracy to the DenseNet feature extractor and smaller but consistent drops to the interaction matrix and to early-layer representations, indicating all three components contribute.
  • Under input perturbations, the paper reports a significant robustness improvement over BERT, so the multi-layer interaction is also claimed to stabilize predictions.

Reading between the lines

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

  • A direct extension is to apply the same stacked-interaction head to other pair-classification tasks, such as retrieval reranking or question-answer verification, where intermediate-layer comparisons may dominate.
  • Because the interaction tensor scales as $n \times m \times d \times L$, long-sequence applications will likely need pooling or sparse sampling over the token-pair grid; the paper does not address that cost.
  • The manuscript leaves unresolved citation placeholders for the robustness-evaluation tool, the MultiNLI ablation dataset, and the optimizer, and never cites the DenseNet paper, so the robustness section and the DenseNet component are not fully traceable from the text alone.
  • The paper reports STS-B as accuracy, which differs from GLUE's official correlation metrics; a reader should verify the quoted gains under the official STS-B protocol before transferring the numbers.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 5 minor

Summary. The paper proposes a Cascaded Interactive Reasoning Network (CIRN), which takes token embeddings from all Transformer layers of a pre-trained language model (e.g., BERT or RoBERTa), computes element-wise interaction tensors between premise and hypothesis tokens at each layer, stacks these tensors along a new dimension, and processes the resulting 4D tensor with a DenseNet before classification. The central claim, stated in the Abstract and Section 5.1, is that CIRN yields consistent average accuracy gains over BERT-base (0.8%), BERT-large (0.7%), RoBERTa-base (1.5%), and RoBERTa-large (0.5%) across ten NLI and sentence-pair datasets.

Significance. If the empirical claims were established, the paper would show that a simple post-hoc interaction module, applied to all layer representations of a pre-trained Transformer, can consistently improve sentence-pair classification accuracy. This would be a useful and easily deployable contribution. However, the current evaluation does not support the claims: the reported averages do not match the table, one key result is an unexplained outlier, the STS-B metric is ambiguous, and no reproducibility information is provided. The architectural novelty is also modest, as element-wise interaction and dense feature extraction are known components; the main potential value lies in the claimed consistent gains, which are not reliably demonstrated.

major comments (5)
  1. [Section 5.1, Table 1] The text claims average accuracy gains of 0.8%, 0.7%, 1.5%, and 0.5% over BERT-base, BERT-large, RoBERTa-base, and RoBERTa-large, respectively, but the Avg column of Table 1 yields differences of 1.0, 0.8, 1.6, and 0.4 percentage points. The paper does not explain this discrepancy, so the headline quantitative claim cannot be reproduced from the table. In addition, no error bars, random seeds, or significance tests are reported, which is a serious problem given that many per-dataset gains are 0.1–0.3 points and some are ties.
  2. [Table 1, RTE column] The RoBERTa-base-CIRN RTE score of 82.5 is a 8.9-point jump over the RoBERTa-base baseline of 73.6, whereas the corresponding RoBERTa-large-CIRN gain is only 0.3 points (84.1 vs. 83.8), and no other model row shows a comparable improvement on any dataset. This single outlier largely drives the claimed 1.5% average gain over RoBERTa-base. The paper offers no explanation for this value; if it is a typo or originates from a different evaluation protocol, the central 'consistent surpass' conclusion is not supported.
  3. [Section 4.1, Table 1] STS-B is a semantic textual similarity task normally evaluated by Pearson or Spearman correlation, but Table 1 reports it as an accuracy number. The paper never specifies how STS-B scores were converted to accuracy (e.g., whether a threshold was used to binarize the continuous similarity labels or whether a regression objective was replaced by classification). Without this information, the STS-B column and the overall averages are not interpretable or comparable with the other datasets.
  4. [Sections 4.2, 4.3, 5.2, and 6] The conclusion states that CIRN shows 'significant improvement in robustness tests', and Section 4.2 refers to robustness evaluations with TextFlint, but no robustness experiments are reported anywhere in the manuscript. Moreover, Sections 4.3 and 5.2 contain placeholder citations '(?)' for the Adadelta optimizer and the MultiNLI dataset, respectively, and the paper does not provide code, data splits, or random seeds. These omissions mean the experiments cannot be reproduced or independently verified.
  5. [Section 3.4] The DenseNet adaptation is underspecified. The interaction tensor Istack is 4D of shape n x m x d x L, but the paper only states that 'convolutional or pooling operations are first applied to reduce dimensionality' without specifying the operation, the output tensor shape, or how the reduced representation is fed into the dense blocks. This is a central architectural component, and the lack of detail prevents the work from being replicated or the design choices from being evaluated.
minor comments (5)
  1. [Abstract / Keywords] The keywords section contains the fragment 'neural language process', which appears to be a truncation of 'natural language processing'.
  2. [Figure 1] The caption uses numbered steps '1) PLM Encoding' through '4) Prediction Layer', but the text in Section 3 refers to the 'left panel' of the figure; the manuscript only contains this single figure, so the 'left panel' phrasing is confusing.
  3. [Section 5.2, Table 2] The ablation description in the text does not match the table. The text says removing the interaction mechanism drops accuracy from 85.1% to 84.6%, which corresponds to row 2, but row 2 is labeled 'Remove first 11 layer', while row 3 is 'Remove interaction matrix'. The mapping between the prose and the table rows needs to be corrected.
  4. [Section 4.3] The choice of Adadelta with a learning rate of 0.5 and a scheduled L2 decay is very unusual for fine-tuning Transformer models and should be justified or cited; this is an important detail for reproducibility.
  5. [References] Several works in the reference list are not cited in the text (e.g., references numbered 38–42), while some citations in the text are incomplete or missing (e.g., reference [3] is listed only as 'arXiv preprint.' and Sections 4.2, 4.3, and 5.2 contain placeholder '(?)' citations).

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: CIRN's reported gains are empirical benchmark evaluations, not derived from fitted inputs, definitions, or load-bearing self-citation.

full rationale

The paper's quantitative claims are experimental accuracies in Table 1 and ablations in Table 2. The model definition (Eqs. 1-10) maps input token embeddings to a classification probability distribution; no equation includes a benchmark accuracy, a baseline result, or the reported average gain. The interaction tensor (Eq. 3) is computed from hidden representations, DenseNet (Eq. 7) transforms that tensor, and classification (Eqs. 8-10) uses standard cross-entropy training; none of these steps encodes the target performance numbers by construction. The hyperparameters in Section 4.3 are conventional tuning choices and are not fitted to the headline deltas. The Section 5.1 claims are therefore empirical evaluations on public benchmarks, not consequences of a fitted parameter renamed as a prediction or of a self-referential definition. The references cited are standard external work; the one citation involving a co-author (Ref. 34) is related work and is not load-bearing for any architectural or uniqueness claim. Internal-consistency problems in Table 1 (for example, the unexplained RTE outlier and the mismatch between the text's average gains and the table's Avg column) are verification and correctness concerns, not circularity. No step reduces by the paper's own equations to its own inputs, so the honest finding is no significant circularity.

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

The method introduces no new physical or conceptual entities. The central claim rests on empirically chosen hyperparameters and on assumptions about how to map diverse benchmark tasks to the NLI format, plus an underspecified DenseNet adaptation.

free parameters (5)
  • DenseNet depth and growth = n=8 layers per block, growth rate g=20
    Manual architecture choices in Section 4.3; affects interaction feature capacity.
  • Dimensionality reduction ratio = eta=0.3, theta=0.5
    Manual settings for initial reduction and transition compression in DenseNet; not justified by ablations.
  • Optimizer hyperparameters = Adadelta rho=0.95, eps=1e-8, initial lr=0.5; SGD lr=3e-4
    Unusual for PLM fine-tuning; chosen by hand in Section 4.3; likely tuned on dev sets.
  • Max sequence lengths = 48 (MNLI), 32 (SNLI), 24 (QQP)
    Ad hoc truncation limits per dataset in Section 4.3; trade-off between speed and information, not analyzed.
  • SNLI mixing ratio = 15% of SNLI added to MNLI training
    Ad hoc training data augmentation in Section 4.3; may improve MNLI results but not ablated.
assumptions (4)
  • domain assumption Pre-trained transformer layer outputs at different depths encode complementary semantic information that is useful for NLI.
    Core motivation of CIRN; Section 3.2 and ablations assume intermediate layers add signal beyond the last layer.
  • ad hoc to paper Element-wise product of premise and hypothesis token embeddings captures semantic interaction relevant to inference.
    Eq. 3 defines I(l) by Hadamard product; no theoretical or empirical justification beyond the reported gains.
  • domain assumption Benchmark datasets and labels are trustworthy and the conversion of non-NLI tasks (QQP, STS-B) to three-way classification is valid.
    Section 4.1 treats paraphrase and similarity tasks as NLI; this mapping is asserted, not validated.
  • ad hoc to paper DenseNet can be adapted to a 4D interaction tensor with the unspecified dimensionality reduction.
    Section 3.4 says 'convolutional or pooling operations are first applied' without specifying which; the implementation is an unverifiable premise.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Boosting Neural Language Inference via Cascaded Interactive Reasoning." pith.science (2026). https://pith.science/paper/IR466JZ2

@misc{pith2026250506607,
  author       = {Pith},
  title        = {Pith review of: Boosting Neural Language Inference via Cascaded Interactive Reasoning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/IR466JZ2}},
  note         = {Machine review of arXiv:2505.06607}
}
read the original abstract

Natural Language Inference (NLI) focuses on ascertaining the logical relationship (entailment, contradiction, or neutral) between a given premise and hypothesis. This task presents significant challenges due to inherent linguistic features such as diverse phrasing, semantic complexity, and contextual nuances. While Pre-trained Language Models (PLMs) built upon the Transformer architecture have yielded substantial advancements in NLI, prevailing methods predominantly utilize representations from the terminal layer. This reliance on final-layer outputs may overlook valuable information encoded in intermediate layers, potentially limiting the capacity to model intricate semantic interactions effectively. Addressing this gap, we introduce the Cascaded Interactive Reasoning Network (CIRN), a novel architecture designed for deeper semantic comprehension in NLI. CIRN implements a hierarchical feature extraction strategy across multiple network depths, operating within an interactive space where cross-sentence information is continuously integrated. This mechanism aims to mimic a process of progressive reasoning, transitioning from surface-level feature matching to uncovering more profound logical and semantic connections between the premise and hypothesis. By systematically mining latent semantic relationships at various representational levels, CIRN facilitates a more thorough understanding of the input pair. Comprehensive evaluations conducted on several standard NLI benchmark datasets reveal consistent performance gains achieved by CIRN over competitive baseline approaches, demonstrating the efficacy of leveraging multi-level interactive features for complex relational reasoning.

Figures

Figures reproduced from arXiv: 2505.06607 by the authors.

Figure 1
Figure 1. The overall architecture of the Cascaded Interactive Reasoning Network [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

42 extracted references · 26 canonical work pages

  1. [1]

    Asynchronous deep interaction network for natural language inference,

    D. Liang, F. Zhang, Q. Zhang, and X.-J. Huang, “Asynchronous deep interaction network for natural language inference,” in Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing, 2019, pp. 2692–2700

    work page 2019

  2. [2]

    Enhanced lstm for natural language inference,

    Q. Chen, X. Zhu, Z. Ling, S. Wei, H. Jiang, and D. Inkpen, “Enhanced lstm for natural language inference,” arXiv preprint arXiv:1609.06038 , 2016

    arXiv 2016

  3. [3]

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

    J. Devlin, M.-W. Chang, K. Lee, and K. Toutanova, “Bert: Pre-training of deep bidirectional transformers for language understanding,” arXiv preprint

    work page

  4. [4]

    Roberta: A robustly optimized bert pretraining approach,

    Y. Liu, M. Ott, N. Goyal, J. Du, M. Joshi, D. Chen, O. Levy, M. Lewis, L. Zettle- moyer, and V. Stoyanov, “Roberta: A robustly optimized bert pretraining approach,” arXiv preprint arXiv:1907.11692 , 2019

    arXiv 1907

  5. [5]

    Semantics-aware bert for language understanding,

    Z. Zhang, Y. Wu, H. Zhao, Z. Li, S. Zhang, X. Zhou, and X. Zhou, “Semantics-aware bert for language understanding,” in AAAI, vol. 34, 2020, pp. 9628–9635

    work page 2020

  6. [6]

    Using prior knowledge to guide bert’s attention in semantic textual matching tasks,

    T. Xia, Y. Wang, Y. Tian, and Y. Chang, “Using prior knowledge to guide bert’s attention in semantic textual matching tasks,” in WWW 2021

    work page 2021

  7. [7]

    What is the jeopardy model? a quasi- synchronous grammar for qa,

    M. Wang, N. A. Smith, and T. Mitamura, “What is the jeopardy model? a quasi- synchronous grammar for qa,” in EMNLP 2007

    work page 2007

  8. [8]

    A large annotated corpus for learning natural language inference,

    S. R. Bowman, G. Angeli, C. Potts, and C. D. Manning, “A large annotated corpus for learning natural language inference,” arXiv preprint arXiv:1508.05326 , 2015

    arXiv 2015

Show all 42 references
  1. [9]

    Supervised learning of universal sentence representations from natural language inference data,

    A. Conneau, D. Kiela, H. Schwenk, L. Barrault, and A. Bordes, “Supervised learning of universal sentence representations from natural language inference data,” arXiv preprint arXiv:1705.02364, 2017

    2017 arXiv

  2. [10]

    Learning to compose task-specific tree structures,

    J. Choi, K. M. Yoo, and S.-g. Lee, “Learning to compose task-specific tree structures,” in Thirty-Second AAAI Conference on Artificial Intelligence , 2018

    2018

  3. [11]

    Bilateral multi-perspective matching for natural language sentences,

    Z. Wang, W. Hamza, and R. Florian, “Bilateral multi-perspective matching for natural language sentences,” arXiv preprint arXiv:1702.03814 , 2017

    2017 arXiv

  4. [12]

    Enhanced-rcnn: an efficient method for learning sentence similarity,

    S. Peng, H. Cui, N. Xie, S. Li, J. Zhang, and X. Li, “Enhanced-rcnn: an efficient method for learning sentence similarity,” in WWW 2020

    2020

  5. [13]

    Enhanced attentive convolutional neural networks for sentence pair modeling,

    S. Xu, E. Shijia, and Y. Xiang, “Enhanced attentive convolutional neural networks for sentence pair modeling,” Expert Systems with Applications

  6. [14]

    Describing multimedia content using attention-based encoder-decoder networks,

    K. Cho, A. Courville, and Y. Bengio, “Describing multimedia content using attention-based encoder-decoder networks,” IEEE Transactions on Multimedia

  7. [15]

    Local and global: temporal question answering via information fusion,

    Y. Liu, M. L. Di Liang, F. Giunchiglia, X. Li, S. Wang, W. Wu, L. Huang, X. Feng, and R. Guan, “Local and global: temporal question answering via information fusion,” in Proceedings of the Thirty-Second International Joint Conference on Artificial Intelligence, 2023, pp. 5141–5149

    2023

  8. [16]

    Deep residual learning for image recognition,

    K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,”

  9. [17]

    Simple and effective text matching with richer alignment features,

    R. Yang, J. Zhang, X. Gao, F. Ji, and H. Chen, “Simple and effective text matching with richer alignment features,” arXiv preprint arXiv:1908.00300 , 2019

    1908 arXiv

  10. [18]

    Distilling structured knowledge for text-based relational reasoning,

    J. Dong, M.-A. Rondeau, and W. L. Hamilton, “Distilling structured knowledge for text-based relational reasoning,” in EMNLP 2020

    2020

  11. [19]

    Xlnet: Generalized autoregressive pretraining for language understanding,

    Z. Yang, Z. Dai, Y. Yang, J. Carbonell, R. R. Salakhutdinov, and Q. V. Le, “Xlnet: Generalized autoregressive pretraining for language understanding,” Advances in neural information processing systems, vol. 32, 2019

    2019

  12. [20]

    Question calibration and multi-hop modeling for temporal question answering,

    C. Xue, D. Liang, P. Wang, and J. Zhang, “Question calibration and multi-hop modeling for temporal question answering,” in AAAI

  13. [21]

    Learning natural language inference using bidirectional lstm model and inner-attention,

    Y. Liu, C. Sun, L. Lin, and X. Wang, “Learning natural language inference using bidirectional lstm model and inner-attention,” arXiv preprint arXiv:1605.09090 . 12 Li et al

    arXiv

  14. [22]

    Local and global: Text matching via syntax graph calibration,

    L. Li, Q. Liao, M. Lai, D. Liang, and S. Liang, “Local and global: Text matching via syntax graph calibration,” in ICASSP 2024-2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

    2024

  15. [23]

    Syntax-bert: Improving pre-trained transformers with syntax trees,

    J. Bai, Y. Wang, Y. Chen, Y. Yang, J. Bai, J. Yu, and Y. Tong, “Syntax-bert: Improving pre-trained transformers with syntax trees,” arXiv preprint, 2021

    2021

  16. [24]

    Glue: A multi-task benchmark and analysis platform for natural language understanding,

    A. Wang, A. Singh, J. Michael, F. Hill, O. Levy, and S. R. Bowman, “Glue: A multi-task benchmark and analysis platform for natural language understanding,” arXiv preprint arXiv:1804.07461 , 2018

    2018 arXiv

  17. [25]

    A continuously growing dataset of sentential paraphrases,

    W. Lan, S. Qiu, H. He, and W. Xu, “A continuously growing dataset of sentential paraphrases,” arXiv preprint arXiv:1708.00391 , 2017

    2017 arXiv

  18. [26]

    Scitail: A textual entailment dataset from science question answering,

    T. Khot, A. Sabharwal, and P. Clark, “Scitail: A textual entailment dataset from science question answering,” in AAAI 2018

    2018

  19. [27]

    Attention is all you need,

    A. Vaswani, N. Shazeer, N. Parmar, A. N. Gomez, L. Kaiser, and I. Polosukhin, “Attention is all you need,” in Advances in neural information processing systems

  20. [28]

    A compare-propagate architecture with align- ment factorization for natural language inference,

    Y. Tay, L. A. Tuan, and S. C. Hui, “A compare-propagate architecture with align- ment factorization for natural language inference,” arXiv preprint arXiv:1801.00102

    arXiv

  21. [29]

    Adaptive multi-attention network incorporating answer information for duplicate question detection,

    D. Liang, F. Zhang, W. Zhang, Q. Zhang, J. Fu, M. Peng, T. Gui, and X. Huang, “Adaptive multi-attention network incorporating answer information for duplicate question detection,” in Proceedings of the 42nd International ACM SIGIR Confer- ence on Research and Development in In...

    2019

  22. [30]

    DABERT: Dual attention enhanced BERT for semantic matching,

    S. Wang, D. Liang, J. Song, Y. Li, and W. Wu, “DABERT: Dual attention enhanced BERT for semantic matching,” in Proceedings of the 29th International Conference on Computational Linguistics , Oct. 2022

    2022

  23. [31]

    Dual path modeling for semantic matching by perceiving subtle conflicts,

    C. Xue, D. Liang, S. Wang, J. Zhang, and W. Wu, “Dual path modeling for semantic matching by perceiving subtle conflicts,” in ICASSP 2023-2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) . IEEE, 2023

    2023

  24. [32]

    Time-aware multiway adaptive fusion network for temporal knowledge graph question answering,

    Y. Liu, D. Liang, F. Fang, S. Wang, W. Wu, and R. Jiang, “Time-aware multiway adaptive fusion network for temporal knowledge graph question answering,” in ICASSP 2023-2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2023, pp. 1–5

    2023

  25. [33]

    Cqg: A simple and effective controlled generation framework for multi-hop question generation,

    Z. Fei, Q. Zhang, T. Gui, D. Liang, S. Wang, W. Wu, and X.-J. Huang, “Cqg: A simple and effective controlled generation framework for multi-hop question generation,” in Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Pape...

    2022

  26. [34]

    Re- solving word vagueness with scenario-guided adapter for natural language inference,

    Y. Liu, M. Li, D. Liang, X. Li, F. Giunchiglia, L. Huang, X. Feng, and R. Guan, “Re- solving word vagueness with scenario-guided adapter for natural language inference,” arXiv preprint arXiv:2405.12434 , 2024

    2024 arXiv

  27. [35]

    Robust lottery tickets for pre-trained language models,

    R. Zheng, R. Bao, Y. Zhou, D. Liang, S. Wang, W. Wu, T. Gui, Q. Zhang, and X. Huang, “Robust lottery tickets for pre-trained language models,” arXiv preprint arXiv:2211.03013, 2022

    2022 arXiv

  28. [36]

    Searching for optimal subword tokenization in cross-domain ner,

    R. Ma, Y. Tan, X. Zhou, X. Chen, D. Liang, S. Wang, W. Wu, T. Gui, and Q. Zhang, “Searching for optimal subword tokenization in cross-domain ner,” arXiv preprint arXiv:2206.03352, 2022

    2022 arXiv

  29. [37]

    Improving semantic matching through dependency-enhanced pre-trained model with adaptive fusion,

    J. Song, D. Liang, R. Li, Y. Li, S. Wang, M. Peng, W. Wu, and Y. Yu, “Improving semantic matching through dependency-enhanced pre-trained model with adaptive fusion,” in Findings of the Association for Computational Linguistics: EMNLP 2022 . Abu Dhabi, United Arab Emirates: As...

    2022

  30. [38]

    Trans- ferring from formal newswire domain with hypernet for twitter pos tagging,

    T. Gui, Q. Zhang, J. Gong, M. Peng, D. Liang, K. Ding, and X.-J. Huang, “Trans- ferring from formal newswire domain with hypernet for twitter pos tagging,” in Boosting Neural Language Inference via Cascaded Interactive Reasoning 13 Proceedings of the 2018 conference on empiric...

    2018

  31. [39]

    Unleashing potential of evidence in knowledge-intensive dialogue generation,

    X. Wu, J. Yang, T. Li, S. Zhang, Y. Du, L. Chai, D. Liang, and Z. Li, “Unleashing potential of evidence in knowledge-intensive dialogue generation,” in ICASSP 2025- 2025 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2025, pp. 1–5

    2025

  32. [40]

    Tablebench: A comprehensive and complex benchmark for table question answering,

    X. Wu, J. Yang, L. Chai, G. Zhang, J. Liu, X. Du, D. Liang, D. Shu, X. Cheng, T. Sun et al., “Tablebench: A comprehensive and complex benchmark for table question answering,” arXiv preprint arXiv:2408.09174 , 2024

    2024 arXiv

  33. [41]

    Comateformer: Combined attention transformer for semantic sentence matching,

    B. Li, D. Liang, and Z. Zhang, “Comateformer: Combined attention transformer for semantic sentence matching,” arXiv preprint arXiv:2412.07220 , 2024

    2024 arXiv

  34. [42]

    S3prompt: Instructing the model with self-calibration, self-recall and self-aggregation to improve in-context learning,

    J. Chen and J. Liu, “S3prompt: Instructing the model with self-calibration, self-recall and self-aggregation to improve in-context learning,” in Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING ...

    2024

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.