Pith. sign in

REVIEW 4 major objections 6 minor 45 references

IM-BERT: Enhancing Robustness of BERT through the Implicit Euler Method

T0 review · 4 major / 6 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 Implicit Euler connections between BERT layers block adversarial attacks without extra parameters or adversarial training.

desk verdict A plausible parameter-free robustness win for BERT, but the theory ignores what the algorithm actually does; worth reviewing, needs a careful revision. read the letter →

arxiv 2505.06889 v1 pith:KVGWCE7P submitted 2025-05-11 cs.CL cs.AI

Mihyeon Kim , Juhyoung Park , Youngbin Kim This is my paper · ORCID
classification cs.CLcs.AI MSC 65L2068T0768T50
keywords IM-BERTimplicitEulermethodadversarialrobustnessODEinterpretationofneuralnetworksAdvGLUEBERTfine-tuningnumericalstabilityfixed-pointiteration
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

What the paper tries to establish is that BERT's vulnerability to adversarial attacks can be reduced by changing the numerical scheme that connects its layers, without adding parameters or using adversarial training. Conceptualizing each layer as one step of an ODE solver, the paper replaces the standard explicit update with an implicit Euler update, whose solution is approximated by a short gradient-descent loop. The theoretical argument shows the implicit scheme is absolutely stable against initial-value perturbations for any step size, while the explicit scheme diverges outside a bounded stability region. On the AdvGLUE benchmark, the resulting IM-BERT model raises average test accuracy from 32.38 to 40.71 over BERT, and it also holds up better when fine-tuned on only 500 or 1,000 training instances. The intended significance is that robustness can be engineered into the forward architecture itself, rather than into the training objective.

What carries the argument

The IM-connection is the object that carries the argument: instead of the residual update $h_t = h_{t-1} + \phi_t(h_{t-1},\theta_t)$, it defines the hidden state implicitly by $h_t = h_{t-1} + \gamma\phi_t(h_t,\theta_t)$ and computes it with $T$ gradient-descent steps on the fixed-point residual, starting from the explicit estimate. The theoretical engine is Proposition 2: for the model equation $dh/dt = \lambda h + \psi(t,x)$ with $\lambda<0$, the error-propagation factor of the implicit Euler method is $1/(1-\gamma\lambda)$, which is smaller than 1 for every step size, so perturbations decay monotonically. This stability result is what the paper cites to justify replacing BERT's monotone inter-layer connections with the IM-connection, and it also motivates the ablation showing that placing the connection in early-to-middle layers helps most.

What would settle it

Measure the residual $\|h_t - h_{t-1} - \phi_t(h_t,\theta_t)\|$ after the $T=5$ gradient steps in a trained IM-BERT model; if that residual is not small relative to the hidden-state norm, the forward pass is far from the implicit solution and the robustness gain cannot be attributed to the absolute stability proved in Proposition 2. A second check is to compare robustness for $T=5$ against an effectively exact solve with a convergence criterion, which would show whether the truncated iteration is what produces the gain.

Watch

Extended reading notes

Core claim

The paper's central claim is that inserting an approximate implicit Euler step between BERT's layers makes the hidden-state dynamics contractive under input perturbations. For the model equation $dh/dt = \lambda h + \psi(t,x)$ with $\lambda<0$, Proposition 2 shows the error between solutions with perturbed and unperturbed initial values decays like $1/(1-\gamma\lambda)^n$ for the implicit Euler method, regardless of the step size $\gamma$; the explicit method, by contrast, requires $|1+\gamma\lambda|<1$ (Proposition 1). The IM-connection realizes the implicit update by minimizing the residual $\|h_t - h_{t-1} - \gamma\phi_t(h_t,\theta_t)\|^2$ through $T$ gradient-descent iterations, initialized at the explicit estimate. Empirically, IM-BERT scores 40.71 versus 32.38 for BERT on AdvGLUE test sets on average, a gain of about 8.3 percentage points, and outperforms BERT by 5.9 points when fine-tuned on low-resource data. The same layer-connection applied to RoBERTa also outperforms several adversarial-training baselines while using only standard fine-tuning.

Load-bearing premise

The stability proof applies to the exact solution of the implicit equation, but the deployed network computes only five gradient-descent updates toward that solution, and the paper does not establish that five updates are close enough for the stability guarantee to hold.

Editorial extensions

If this is right

  • If the central claim is right, robustness against adversarial inputs can be improved for any BERT-like encoder by changing only the forward propagation, with no new trainable parameters and no adversarial training.
  • The stability argument does not depend on the step size, so the method should require less attack-specific hyperparameter tuning than training-based defenses such as SMART or FreeLB.
  • In low-resource settings, where fine-tuning overfits and adversarial training is hardest to apply, the same IM-connection still gives a measurable accuracy gain over BERT on AdvGLUE.
  • The paper's placement ablation implies that inserting IM-connections only in middle layers can capture most of the robustness gain while reducing the computational cost several-fold compared to applying it everywhere.

Reading between the lines

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

  • A natural extension would be to apply the same implicit-layer idea to embedding layers or attention outputs, not just inter-layer connections; the paper does not test those positions, and the fixed-point contraction may behave differently there.
  • If the mechanism is truly contraction of the implicit map, robustness should improve monotonically with the iteration count $T$ until the fixed point is reached and then plateau; the reported TextFooler results (T=1 to T=15) are consistent with that trend but do not explicitly demonstrate the plateau.
  • Because IM-connection changes only forward propagation, it is in principle composable with adversarial training; whether the two defenses combine additively or redundantly is an untested question the paper leaves open.
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

4 major / 6 minor

Summary. The paper proposes IM-BERT, a modification of BERT that replaces the monotone inter-layer connections with an `IM-connection` modeled on the implicit Euler method. The authors analyze the absolute stability of explicit versus implicit Euler solvers on a linear model equation, prove that the implicit method is absolutely stable regardless of step size, and then implement the implicit update by unrolled gradient descent on the fixed-point residual. Experiments on the AdvGLUE benchmark report that IM-BERT improves over BERT by about 8.3 percentage points on average test accuracy and by about 5.9 points in low-resource settings, without adding parameters or using adversarial training. The paper also reports ablations on layer placement, iteration count, and comparisons with adversarial-training baselines on BERT and RoBERTa.

Significance. If the claimed robustness mechanism were established, the contribution would be valuable: a simple architectural intervention that improves adversarial robustness of pre-trained language models without adversarial training or extra parameters. The paper contains a standard and correct stability analysis for the exact implicit Euler method, a clearly specified architectural change, and experiments on a widely used benchmark, including low-resource settings and FLOPs accounting. However, the central link between the theory and the implemented algorithm is currently broken, and the headline empirical claims lack statistical support. The strengths are the clean theoretical framing, the architectural simplicity, and the breadth of experiments; the weakness is that the theory does not apply to the actual implementation as written.

major comments (4)
  1. [Section 3.4, Algorithm 1 vs Eq. (6) and Eq. (9)] Algorithm 1 does not implement the implicit Euler equation used in Proposition 2. Eq. (6) defines h_t = h_{t-1} + γ φ_t(h_t), and Eq. (9) minimizes ||x - h_{t-1} - γ φ_t(x)||^2. However, Algorithm 1 line 3 computes loss_i = ||h^i_t - h_{t-1} - φ_t(h^i_t)||^2, omitting the factor γ. Even if the gradient descent converges, it converges to the fixed point of x = h_{t-1} + φ_t(x), which is an implicit Euler step with step size 1, not step size γ. The parameter γ=0.1 is thus only a gradient-descent step size in the implemented algorithm, not the ODE step size. Consequently, the absolute-stability guarantee of Proposition 2 does not apply to the architecture that was actually trained and evaluated.
  2. [Section 3.4 and Limitation] The T-step truncation is unanalyzed. Proposition 2 concerns the exact fixed point of the implicit Euler equation, while Algorithm 1 returns the iterate after T=5 unrolled gradient-descent steps. The paper provides no Lipschitz or contractivity bound on the residual map and no error bound showing that this iterate is close enough to the fixed point to inherit the stability property. The Limitation section acknowledges only the time cost, not this approximation gap, so the empirical robustness improvement is currently attributed to a theoretical guarantee that does not cover the implemented computation.
  3. [Section 3.3, Eq. (7)] The stability analysis assumes the model equation dh/dt = λh(t) + ψ(t,x) with λ<0, but the paper does not verify that BERT layers satisfy λ<0 or even discuss the plausibility of this assumption for layer dynamics. Without such a check, Propositions 1 and 2 do not directly transfer to the actual BERT layers. A concrete empirical test of this assumption, or at least a careful discussion of its scope, is needed to make the theoretical argument load-bearing for the proposed architecture.
  4. [Table 1 and Section 4.1] The headline test-set claim of an 8.3 percentage point average improvement over BERT is reported without error bars or significance tests. While Section 4.3 gives three-run averages with standard deviations for SST-2, the main results in Table 1 and the low-resource results in Table 2 are point estimates. Given the per-task variability visible in Table 3, the reader cannot assess whether the average improvements are statistically meaningful. The comparisons with adversarial-training baselines use numbers from other papers under potentially different training conditions, which should be stated explicitly.
minor comments (6)
  1. [Section 3.4, Eq. (9)] The variable x is reused in Eq. (9) as the optimization variable and as the initial-value input of the ODE; this obscures the derivation and should be renamed.
  2. [Algorithm 1] The residual in line 3 should include the factor γ to make the algorithm consistent with Eq. (9); if the omission is intentional, the fixed point of the iteration should be stated and analyzed.
  3. [Introduction] The reference in the first paragraph that appears as "?" should be completed.
  4. [Section 4.2] The phrase "advertising attacks" should read "adversarial attacks."
  5. [Table 3] The row labels for layer groups (e.g., "Layer (1-3)") are missing the model name or a model indicator; the table is hard to parse without knowing which column corresponds to which model.
  6. [Section 3.4, Figure 1] The figure caption is not self-contained; it should explain the meaning of the red and blue connections and the difference between BERT, EX-BERT, and IM-BERT in the figure itself.

Circularity Check

0 steps flagged · score 0.0 of 10

No material circularity: the stability theorem is a standard independent result and the AdvGLUE gains are empirical, not derived from fitted inputs.

full rationale

The paper's derivation chain is: Eq. (4) frames the network as an initial-value ODE; Eqs. (5)–(6) define explicit and implicit Euler discretizations; Propositions 1–2 prove standard absolute-stability conditions on the model equation dh/dt = λh + ψ with λ < 0; Section 3.4 proposes Algorithm 1 to approximate the implicit fixed point; and Section 4 measures AdvGLUE accuracy. None of these steps defines a derived quantity in terms of a target quantity, fits a parameter and then renames it as a prediction, or imports a load-bearing uniqueness claim from the authors' own prior work. The closest concern is that Algorithm 1's fixed-point residual omits the γ from Eq. (6) and the T=5 iterate is not formally linked to the exact implicit solution; that is an implementation/theory mismatch and a correctness risk, not circularity, because Proposition 2 does not use Algorithm 1's outputs as inputs. Hyperparameters T, γ, and layer placement are selected and ablated on benchmarks, but the paper reports them as chosen configuration settings rather than as predictions forced by the theory. There are no relevant self-citations: the cited IE-skips work (Li et al., 2020) is by other authors and is genuine external provenance. The central robustness claim is therefore not equivalent to its inputs by construction.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central implementation (IM-connection) relies on two unverified premises: the negative-eigenvalue model for BERT layers and the sufficiency of a few gradient steps to reach the implicit solution. Hyperparameters T, gamma, and layer placement are tuned on the evaluation benchmark.

free parameters (3)
  • gradient descent step size gamma = 0.1
    Set in Algorithm 1 and experiments; no sensitivity analysis reported.
  • iteration count T = 5 (main), 10, 15 explored
    Selected based on validation performance in Appendix A.1; the paper reports different T for low-resource experiments.
  • IM-connection layer placement = Layers 4-6 for BERT main; others in ablations
    Chosen after ablations (Table 3) because middle layers performed best; this is model selection on the same benchmark used for evaluation.
assumptions (4)
  • domain assumption BERT layer dynamics can be modeled as dh/dt = phi(h(t), theta(t)) with the model equation lambda < 0 for stability analysis.
    Sections 3.1 and 3.3. The layer is treated as an ODE right-hand side; the negative-eigenvalue assumption is not verified for BERT layers.
  • ad hoc to paper Exact solution of the implicit Euler equation (Eq. 6) is well-approximated by T=5 gradient descent steps (Algorithm 1).
    No error bound or contraction analysis provided; Proposition 2 applies to the exact implicit solve, not to the truncated gradient descent approximation.
  • domain assumption Adversarial attacks on text can be treated as initial-value perturbations eta of the hidden-state ODE.
    Section 3.3; the connection between discrete token perturbations and continuous hidden-state perturbations is assumed, not established.
  • standard math Standard numerical analysis of absolute stability (Definitions 1-2, Propositions 1-2) applies.
    Section 3.3; classical results for the scalar model equation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of IM-BERT: Enhancing Robustness of BERT through the Implicit Euler Method." pith.science (2026). https://pith.science/paper/KVGWCE7P

@misc{pith2026250506889,
  author       = {Pith},
  title        = {Pith review of: IM-BERT: Enhancing Robustness of BERT through the Implicit Euler Method},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KVGWCE7P}},
  note         = {Machine review of arXiv:2505.06889}
}
read the original abstract

Pre-trained Language Models (PLMs) have achieved remarkable performance on diverse NLP tasks through pre-training and fine-tuning. However, fine-tuning the model with a large number of parameters on limited downstream datasets often leads to vulnerability to adversarial attacks, causing overfitting of the model on standard datasets. To address these issues, we propose IM-BERT from the perspective of a dynamic system by conceptualizing a layer of BERT as a solution of Ordinary Differential Equations (ODEs). Under the situation of initial value perturbation, we analyze the numerical stability of two main numerical ODE solvers: the explicit and implicit Euler approaches. Based on these analyses, we introduce a numerically robust IM-connection incorporating BERT's layers. This strategy enhances the robustness of PLMs against adversarial attacks, even in low-resource scenarios, without introducing additional parameters or adversarial training strategies. Experimental results on the adversarial GLUE (AdvGLUE) dataset validate the robustness of IM-BERT under various conditions. Compared to the original BERT, IM-BERT exhibits a performance improvement of approximately 8.3\%p on the AdvGLUE dataset. Furthermore, in low-resource scenarios, IM-BERT outperforms BERT by achieving 5.9\%p higher accuracy.

Figures

Figures reproduced from arXiv: 2505.06889 by the authors.

Figure 1
Figure 1. The l-th layer in each architecture The perturbed inputs pass through the layer to estimate the hidden states. Blue represents the corrected hidden state, while red indicates the opposite. Blue connections represent explicit residual connections, while red connections represent IM-connections. (a) BERT (b) EX-BERT In the layer, the layers are connected with the explicit method. (c) IM-BERT The hidden states taken fr… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

45 extracted references · 37 canonical work pages

  1. [1]

    Armen Aghajanyan, Akshat Shrivastava, Anchit Gupta, Naman Goyal, Luke Zettlemoyer, and Sonal Gupta. 2021. Better fine-tuning by reducing representational collapse. In International Conference on Learning Representations

    work page 2021

  2. [2]

    Kendall E Atkinson and Weimin Han. 1993. Elementary numerical analysis. Wiley New York

    work page 1993

  3. [3]

    Jimmy Lei Ba, Jamie Ryan Kiros, and Geoffrey E Hinton. 2016. Layer normalization. arXiv preprint arXiv:1607.06450

    arXiv 2016

  4. [4]

    Aaron Baier-Reinio and Hans De Sterck. 2020. N-ode transformer: A depth-adaptive variant of the transformer using neural ordinary differential equations. arXiv preprint arXiv:2010.11358

    arXiv 2020

  5. [5]

    Martin Braun and Martin Golubitsky. 1983. Differential equations and their applications, volume 2. Springer

    work page 1983

  6. [6]

    Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al. 2020. Language models are few-shot learners. Advances in Neural Information Processing Systems, 33:1877--1901

    2020

  7. [7]

    Ricky TQ Chen, Yulia Rubanova, Jesse Bettencourt, and David K Duvenaud. 2018. Neural ordinary differential equations. Advances in Neural Information Processing Systems, 31:6571--6583

    work page 2018

  8. [8]

    Chengyu Dong, Liyuan Liu, Zichao Li, and Jingbo Shang. 2020. Towards adaptive residual network training: A neural-ode perspective. In Proceedings of the 37th International Conference on Machine Learning, pages 2616--2626

    work page 2020

Show all 45 references
  1. [9]

    Eldad Haber and Lars Ruthotto. 2017. Stable architectures for deep neural networks. Inverse problems, 34(1):014004

    2017

  2. [10]

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. 2016. Deep residual learning for image recognition. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 770--778

    2016

  3. [11]

    Haoming Jiang, Pengcheng He, Weizhu Chen, Xiaodong Liu, Jianfeng Gao, and Tuo Zhao. 2020. Smart: Robust and efficient fine-tuning for pre-trained natural language models through principled regularized optimization. In Proceedings of the 58th Annual Meeting of the Association f...

    2020

  4. [12]

    Di Jin, Zhijing Jin, Joey Tianyi Zhou, and Peter Szolovits. 2020. Is bert really robust? a strong baseline for natural language attack on text classification and entailment. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 34, pages 8018--8025

    2020

  5. [13]

    Kenji Kawaguchi. 2020. On the theory of implicit deep learning: Global convergence with implicit layers. In International Conference on Learning Representations

    2020

  6. [14]

    Jacob Devlin Ming-Wei Chang Kenton and Lee Kristina Toutanova. 2019. Bert: Pre-training of deep bidirectional transformers for language understanding. In Conference of the North American Chapter of the Association for Computational Linguistics, pages 4171--4186

    2019

  7. [15]

    Diederik P Kingma and Jimmy Ba. 2015. Adam: A method for stochastic optimization. International Conference on Learning Representations

    2015

  8. [16]

    Bei Li, Quan Du, Tao Zhou, Yi Jing, Shuhan Zhou, Xin Zeng, Tong Xiao, JingBo Zhu, Xuebo Liu, and Min Zhang. 2022. Ode transformer: An ordinary differential equation-inspired model for sequence generation. In Proceedings of the 60th Annual Meeting of the Association for Computa...

    2022

  9. [17]

    Mingjie Li, Lingshen He, and Zhouchen Lin. 2020. Implicit euler skip connections: Enhancing adversarial robustness via numerical stability. In Proceedings of the International Conference on Machine Learning, pages 5874--5883

    2020

  10. [18]

    Jieyu Lin, Jiajie Zou, and Nai Ding. 2021. Using adversarial attacks to reveal the statistical bias in machine reading comprehension models. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference o...

    2021

  11. [19]

    Yiping Lu, Zhuohan Li, Di He, Zhiqing Sun, Bin Dong, Tao Qin, Liwei Wang, and Tie-yan Liu. 2020. Understanding and improving transformer from a multi-particle dynamic system point of view. In ICLR 2020 Workshop on Integration of Deep Neural Models and Differential Equations

    2020

  12. [20]

    Yiping Lu, Aoxiao Zhong, Quanzheng Li, and Bin Dong. 2018. Beyond finite layer neural networks: Bridging deep architectures and numerical differential equations. In Proceedings of the 35th International Conference on Machine Learning, pages 3276--3285

    2018

  13. [21]

    Andrew Maas, Raymond E Daly, Peter T Pham, Dan Huang, Andrew Y Ng, and Christopher Potts. 2011. Learning word vectors for sentiment analysis. In Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies, pages 142--150

    2011

  14. [22]

    Bonan Min, Hayley Ross, Elior Sulem, Amir Pouran Ben Veyseh, Thien Huu Nguyen, Oscar Sainz, Eneko Agirre, Ilana Heintz, and Dan Roth. 2021. Recent advances in natural language processing via large pre-trained language models: A survey. ACM Computing Surveys

    2021

  15. [23]

    Yixin Nie, Adina Williams, Emily Dinan, Mohit Bansal, Jason Weston, and Douwe Kiela. 2020. Adversarial nli: A new benchmark for natural language understanding. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 4885--4901

    2020

  16. [24]

    Colin Raffel, Noam Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, and Peter J Liu. 2020. Exploring the limits of transfer learning with a unified text-to-text transformer. The Journal of Machine Learning Research, 21(1):5485--5551

    2020

  17. [25]

    Viktor Reshniak and Clayton G Webster. 2020. Robust learning with implicit residual networks. Machine Learning and Knowledge Extraction, 3(1):34--55

    2020

  18. [26]

    Marco Tulio Ribeiro, Tongshuang Wu, Carlos Guestrin, and Sameer Singh. 2020. Beyond accuracy: Behavioral testing of nlp models with checklist. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 4902--4912

    2020

  19. [27]

    Robin M Schmidt. 2019. Recurrent neural networks (rnns): A gentle introduction and overview. arXiv preprint arXiv:1912.05911

    2019 arXiv

  20. [28]

    Jiawei Shen, Zhuoyan Li, Lei Yu, Gui-Song Xia, and Wen Yang. 2020. Implicit euler ode networks for single-image dehazing. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, pages 218--219

    2020

  21. [29]

    Zhiqing Sun, Hongkun Yu, Xiaodan Song, Renjie Liu, Yiming Yang, and Denny Zhou. 2020. Mobilebert: a compact task-agnostic bert for resource-limited devices. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 2158--2170

    2020

  22. [30]

    Florian Tramer and Dan Boneh. 2019. Adversarial training and robustness for multiple perturbations. Advances in neural information processing systems, 32

    2019

  23. [31]

    Alex Wang, Amanpreet Singh, Julian Michael, Felix Hill, Omer Levy, and Samuel R Bowman. 2018. Glue: A multi-task benchmark and analysis platform for natural language understanding. Proceedings of the 2018 EMNLP Workshop BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP

    2018

  24. [32]

    Boxin Wang, Shuohang Wang, Yu Cheng, Zhe Gan, Ruoxi Jia, Bo Li, and Jingjing Liu. 2021 a . Infobert: Improving robustness of language models from an information theoretic perspective. In International Conference on Learning Representations

    2021

  25. [33]

    Boxin Wang, Chejian Xu, Shuohang Wang, Zhe Gan, Yu Cheng, Jianfeng Gao, Ahmed Hassan Awadallah, and Bo Li. 2021 b . Adversarial glue: A multi-task benchmark for robustness evaluation of language models. In Thirty-fifth Conference on Neural Information Processing Systems Datase...

    2021

  26. [34]

    Hongqiu Wu, Yongxiang Liu, Hanwen Shi, Min Zhang, et al. 2022. Toward adversarial training on contextualized language representation. In International Conference on Learning Representations

    2022

  27. [35]

    Yibo Yang, Jianlong Wu, Hongyang Li, Xia Li, Tiancheng Shen, and Zhouchen Lin. 2020. Dynamical system inspired adaptive time stepping controller for residual network families. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 34, pages 6648--6655

    2020

  28. [36]

    Penghang Yin, Minh Pham, Adam Oberman, and Stanley Osher. 2018. Stochastic backward euler: an implicit gradient descent algorithm for k-means clustering. Journal of Scientific Computing, 77:1133--1146

    2018

  29. [37]

    Lifan Yuan, Yichi Zhang, Yangyi Chen, and Wei Wei. 2021. Bridge the gap between cv and nlp! a gradient-based textual adversarial attack framework. arXiv preprint arXiv:2110.15317

    2021 arXiv

  30. [38]

    Jing Zhang, Peng Zhang, Baiwen Kong, Junqiu Wei, and Xin Jiang. 2021. Continuous self-attention models with neural ode networks. In Thirty-Fifth AAAI Conference on Artificial Intelligence, volume 35, pages 14393--14401

    2021

  31. [39]

    Jingfeng Zhang, Bo Han, Laura Wynter, Bryan Kian Hsiang Low, and Mohan Kankanhalli. 2019. Towards robust resnet: A small step but a giant leap. In Proceedings of the Twenty-Eighth International Joint Conference on Artifcial Intelligence, pages 4285--4291

    2019

  32. [40]

    Xingcheng Zhang, Zhizhong Li, Chen Change Loy, and Dahua Lin. 2017. Polynet: A pursuit of structural diversity in very deep networks. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 718--726

    2017

  33. [41]

    Rui Zheng, Rong Bao, Qin Liu, Tao Gui, Qi Zhang, Xuan-Jing Huang, Rui Xie, and Wei Wu. 2022. Plugat: A plug and play module to defend against textual adversarial attack. In Proceedings of the 29th International Conference on Computational Linguistics, pages 2873--2882

    2022

  34. [42]

    Chen Zhu, Yu Cheng, Zhe Gan, Siqi Sun, Tom Goldstein, and Jingjing Liu. 2020. Freelb: Enhanced adversarial training for natural language understanding. In International Conference on Learning Representations

    2020

  35. [43]

    Mai Zhu, Bo Chang, and Chong Fu. 2023. Convolutional neural networks combined with runge-kutta methods. Neural Computing and Applications, 35(2):1629--1643

    2023

  36. [44]

    URL: " 'urlintro :=

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

  37. [45]

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