Pith. sign in

REVIEW 5 major objections 6 minor 33 references

Loss-Aware Curriculum Learning for Chinese Grammatical Error Correction

T0 review · 5 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read The paper argues that training Chinese grammatical error correction models on example pairs sorted from easy to hard by model loss, and then reweighting individual examples by a dropout-based confidence estimate, consistently improves…

desk verdict Plausible curriculum learning idea for CGEC, but the instance-level mechanism is undefined, making the reported gains unverifiable as written. read the letter →

arxiv 2501.00334 v1 pith:WDJICMGH submitted 2024-12-31 cs.CL cs.AI

Ding Zhang , Yangning Li , Lichen Bai , Hao Zhang , Yinghui Li , Haiye Lin , Hai-Tao Zheng , Xin Su
show 1 more author
Zifei Shan
This is my paper · ORCID
classification cs.CLcs.AI
keywords Chinesegrammaticalerrorcorrectioncurriculumlearningloss-awaredifficultybatch-levelinstance-levelreweightingMonteCarlodropoutpre-trainedlanguagemodels
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

Chinese grammatical error correction systems usually train on every sentence equally, even though some sentences are much harder to fix than others. This paper proposes a two-level curriculum: first, sort training pairs by a model-loss difficulty score and feed them in batches from easy to hard; then, inside each batch, reweight instances according to a dropout-based confidence estimate. The authors test the recipe on BART, mT5, and SynGEC and report consistent F0.5 gains on NLPCC-test and MuCGEC-dev. A sympathetic reading is that difficulty-aware training alone, with no change to model architecture or data, can push CGEC performance upward.

What carries the argument

The paper's central object is a two-level curriculum driven by a loss-based difficulty score. At batch level, each sentence pair $(x^i,y^i)$ receives a difficulty score $d(x^i,y^i)$ equal to its cross-entropy loss under a trained CGEC model, normalized to $(0,1]$ by a cumulative density function; a competence schedule $C(t)=\min(1,\sqrt[k]{t(1-c_0^k)/T + c_0^k})$ determines which fraction of the difficulty-sorted data is exposed at training step $t$. At instance level, $Q=5$ Monte Carlo dropout forward passes yield $Q$ conditional probabilities per sentence and token; the variance of these probabilities acts as a confidence measure that reweights the loss so hard instances contribute more. The negative log-likelihood loss in Equation (1) is the quantity both levels operate on.

What would settle it

Sort a held-out set by a trained model's per-sentence loss and check whether correction accuracy improves monotonically as loss decreases; then retrain the whole model with the curriculum reversed, hard to easy. If the loss–difficulty correlation is weak or the reversed curriculum produces the same F0.5 gains, the easy-to-hard ordering is not the mechanism driving the reported improvement.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that a loss-aware, two-granularity curriculum—batch-level ordering plus instance-level reweighting—improves CGEC models independent of the underlying architecture. Reported F0.5 gains on NLPCC-test are 1.17 points for BART, 1.11 for mT5, and 0.73 for SynGEC; on MuCGEC-dev they are 0.97, 0.55, and 0.87 respectively. The authors attribute the gains to gradually shifting model attention toward hard examples, and the ablation study shows both curriculum levels contribute.

Load-bearing premise

The curriculum ordering stands or falls on the assumption that the model's per-sentence training loss is a reliable, stable measure of how hard a sentence is to correct, so that low-loss sentences are genuinely easy and high-loss sentences genuinely hard.

Editorial extensions

If this is right

  • On NLPCC-test, adding the curriculum to BART raises F0.5 from 44.64 to 45.81, to mT5 from 42.81 to 43.92, and to SynGEC from 45.32 to 46.05.
  • On MuCGEC-dev, the same recipe raises mT5 from 36.87 to 37.42, BART from 37.62 to 38.59, and SynGEC from 38.31 to 39.18.
  • Ablations on BART show the combined gain (+1.17) is larger than batch-level only (+0.38) or instance-level only (+0.64), so the two granularities are complementary.
  • The competence-growth parameter $k$ has a sweet spot: F0.5 peaks at $k=2$ on both NLPCC-test and MuCGEC-dev, and larger $k$ degrades the gain, consistent with overfitting hard examples.
  • Figure 2 indicates gains appear across all difficulty intervals, not only the hardest sentences.

Reading between the lines

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

  • The paper does not test whether the loss-based difficulty ranking is stable across random seeds; if rankings fluctuate, a cheaper variant would be to compute difficulty once from a small proxy model or to smooth scores over an ensemble of checkpoints.
  • The same two-level recipe should transfer to English grammatical error correction or spell-checking, since neither the CDF-normalized difficulty score nor the dropout-confidence reweighting is Chinese-specific.
  • A direct extension would be to replace the single-score CDF ordering with token-level or error-type-aware curricula, since Figure 2 shows hard sentences benefit most; error-type annotation could make the easy-to-hard path linguistically interpretable.
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 / 6 minor

Summary. The paper proposes a multi-granularity curriculum learning framework for Chinese grammatical error correction (CGEC). Batch-level curriculum learning scores training instances by the cross-entropy loss of a trained CGEC model and presents them in an easy-to-hard order controlled by a competence function. Instance-level curriculum learning is described as using Monte Carlo dropout sampling to obtain conditional probabilities and then adjusting loss weights based on the variance of those probabilities. The method is applied to BART, mT5, and SynGEC, and results are reported on NLPCC-test and MuCGEC-dev, with ablations, a parameter study, and a difficulty-interval analysis. The central claim is that this training-only, architecture-agnostic procedure yields consistent F0.5 improvements of roughly 0.5 to 1.2 points over baselines.

Significance. If the improvements were established, the proposed curriculum would be a useful, model-agnostic training strategy for CGEC and would connect curriculum learning more tightly to error-correction difficulty. The paper's intended contribution is plausible and the experimental design covers multiple base architectures, which is a strength. However, as written, the instance-level component is not specified as a reproducible mechanism, the evaluation setup includes selection of the hyperparameter k on the test sets, and no statistical significance evidence accompanies the reported seed averages. These issues currently prevent the claimed results from being verified, so the significance of the contribution is not yet demonstrated.

major comments (5)
  1. [II-C] The Instance-Level Curriculum Learning mechanism is not defined. The text states that Monte Carlo dropout sampling yields Q conditional probabilities per instance and that the variance of these probabilities reflects model confidence, but it never gives the equation that maps these probabilities or variances to modified training losses. The abstract and conclusion say the loss function is regulated, but no formula, normalization, temperature, or threshold is provided. Because the main results in Tables III and IV and the +0.64 F0.5 ablation in Table V all depend on this component, the central claim cannot be independently checked or reproduced from the paper as written.
  2. [III-F, Table VI] The hyperparameter k is selected by scanning F0.5 directly on the NLPCC-test and MuCGEC-dev datasets. The reported CL(BART) scores are therefore the best k values on the evaluation sets, which introduces selection bias into the claimed improvements. The authors should either use a held-out validation split for selecting k and report test results only for the selected value, or report results across all k values on a separate test set. Without this, the magnitudes of the reported gains are not reliable evidence for the method's effectiveness.
  3. [II-B1] The difficulty-scoring model is underspecified. The paper says 'Given a trained CGEC model and a dataset... we can acquire the difficulty of each instance by calculating the cross-entropy loss function,' but it does not state which architecture, which checkpoint, or which training data produced this scorer. This ambiguity matters because if each of the three base models is used to score its own training data, the batch-level curriculum differs per model and the comparison is confounded; the authors should fix a scoring model and report its provenance, or alternatively demonstrate that the loss-based difficulty ordering is stable across initializations and architectures.
  4. [III-D, Table III and Table IV] The claim that the method 'achieves consistent improvements with a significant margin over all baselines' is not supported by statistical evidence. The paper states that results are averages of five distinct experiments with different random seeds, but no standard deviations, confidence intervals, or significance tests are reported. Given that the observed differences are on the order of 0.5 to 1.2 F0.5, run-to-run variance in fine-tuning could substantially affect the conclusions, so the significance claim should be substantiated or softened.
  5. [I, III-D] The most closely related baselines are absent from the experimental comparison. The introduction cites [10] (mixed-grained weighted training) and [11] (data weighted training strategies) as relevant prior work on addressing instance difficulty in GEC, but neither is included in Tables III or IV. Since these methods directly weight training data based on difficulty, they are the natural competitors for the proposed approach, and their omission undermines the 'all baselines' claim in Section III-D.
minor comments (6)
  1. [Abstract and throughout] The abbreviation 'PLMS' appears in the abstract; it should be 'PLMs' (pre-trained language models). This typo recurs in the introduction and should be corrected.
  2. [II-C] Monte Carlo dropout sampling is introduced without a citation. Please provide a reference (e.g., Gal and Ghahramani, 2016) and specify how dropout is applied during inference and which layers are affected.
  3. [II-B2, Eq. (3)] The notation for the k-th root is unclear; 'k√' should be typeset as \(\sqrt[k]{...}\) or written in an equivalent unambiguous form.
  4. [III-F, Table VI] The table caption and text contain the phrase 'valuation of k'; this should be 'value of k'.
  5. [III-G, Figure 2] The figure is not referenced in the text at the point where the difficulty-interval analysis is introduced. Additionally, the y-axis label 'Improvements' should be quantified as ΔF0.5 with units specified.
  6. [III-C] The learning rate is printed as '5e−5' with an ambiguous minus sign; please use standard scientific notation, e.g., \(5\times 10^{-5}\).

Circularity Check

1 steps flagged · score 4.0 of 10

Reported gains are partly selection-maximized: the hyperparameter k is tuned on the same test sets used for evaluation.

  1. fitted input called prediction [Section III-F, Table VI (and Tables III/IV)]
    "In this section, to investigate the impact of different k, we perform extensive experiments by varying the valuation of k on the NLPCC-test dataset. Table VI shows that when the value of k reaches a certain value, the performance of the model does not improve anymore. Consequently, it is essential to choose the optimal value of k, although there are consistent improvements when using BART as the PLM at all values of k from 1 to 5."

    The reported CL(BART) scores in Tables III and IV (45.81 on NLPCC-test and 38.59 on MuCGEC-dev) match exactly the best-k rows in Table VI for k=2 on the same two test sets. Because the parameter study is performed directly on the evaluation sets, the claimed 'consistent improvements with a significant margin' are selection maxima over k rather than out-of-sample predictions. The evaluation loop is thereby partially closed: the same data used to choose the hyperparameter is then used to certify the method's improvement, so the headline margin is inflated by construction. The qualitative claim is not entirely forced because k=1 through k=5 all improve over BART, but the specific reported margin is selected.

full rationale

The derivation chain is not formally circular in the sense of Eq. X = Eq. Y by construction: the difficulty scores are defined by a trained CGEC model's loss, but the final evaluation uses held-out NLPCC-test and MuCGEC-dev F0.5 scores, so the central improvement claim is an empirical result rather than a tautology. The batch-level competence schedule is adopted from external prior work (Platanios et al.), and no load-bearing self-citation or uniqueness theorem is invoked. The main circularity concern is the hyperparameter selection: Section III-F tunes k on the two test sets, and the CL(BART) results in Tables III/IV are exactly the best-k values from Table VI, so the reported margin is partly a fitted maximum. A separate, non-circular reproducibility gap is that Section II-C never specifies the loss-weighting formula for Instance-Level CL, making the +0.64 F0.5 ablation unverifiable; this affects correctness and replicability but does not by itself make the derivation circular. Overall, the central claim retains independent content because even the worst k improves over the baseline, but the specific significant margin is compromised by test-set selection, warranting a moderate circularity score.

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

No new physical or conceptual entity is introduced. The framework is composed of known ingredients, but it rests on several unvalidated empirical assumptions about loss, dropout variance, and curriculum ordering, plus unreported hyperparameters.

free parameters (4)
  • k (competence growth coefficient) = 2 (selected on NLPCC-test and MuCGEC-dev in Table VI)
    Growth coefficient in Eq. 3; reported main results use the best-scoring value, a test-set-fitted choice.
  • c0 (initial competence) = not reported
    Initial competence in Eq. 3; no value or selection procedure is provided.
  • T (curriculum length) = not reported
    Hyperparameter determining the length of the Batch-Level curriculum in Eq. 3; not specified.
  • Q (Monte Carlo dropout samples) = 5
    Number of MC dropout samples used to estimate confidence; no sensitivity analysis is given.
assumptions (5)
  • domain assumption Loss of a trained CGEC model is a monotone measure of correction difficulty.
    Section II-B1 states that lower loss implies easier correction and higher loss implies harder correction; no independent validation is supplied.
  • ad hoc to paper Monte Carlo dropout variance over Q=5 samples reflects model confidence and instance/token difficulty.
    Section II-C uses this as the basis of Instance-Level CL without evidence that five samples give stable or meaningful uncertainty estimates.
  • domain assumption The competence-based curriculum schedule from Platanios et al. transfers to CGEC.
    Eq. 3 is imported from prior NMT work without adaptation or validation for Chinese grammatical error correction.
  • domain assumption Easy-to-hard ordering is beneficial for CGEC training.
    This is the core premise of curriculum learning and is not separately established for this task beyond the reported gains.
  • domain assumption SynGEC optimization hyperparameters are valid for BART and mT5.
    Section III-C says the authors adopt hyperparameters of SynGEC to guide optimization of all models, with no per-model tuning or justification.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Loss-Aware Curriculum Learning for Chinese Grammatical Error Correction." pith.science (2026). https://pith.science/paper/WDJICMGH

@misc{pith2026250100334,
  author       = {Pith},
  title        = {Pith review of: Loss-Aware Curriculum Learning for Chinese Grammatical Error Correction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WDJICMGH}},
  note         = {Machine review of arXiv:2501.00334}
}
read the original abstract

Chinese grammatical error correction (CGEC) aims to detect and correct errors in the input Chinese sentences. Recently, Pre-trained Language Models (PLMS) have been employed to improve the performance. However, current approaches ignore that correction difficulty varies across different instances and treat these samples equally, enhancing the challenge of model learning. To address this problem, we propose a multi-granularity Curriculum Learning (CL) framework. Specifically, we first calculate the correction difficulty of these samples and feed them into the model from easy to hard batch by batch. Then Instance-Level CL is employed to help the model optimize in the appropriate direction automatically by regulating the loss function. Extensive experimental results and comprehensive analyses of various datasets prove the effectiveness of our method.

Figures

Figures reproduced from arXiv: 2501.00334 by the authors.

Figure 1
Figure 1. An Overview of Loss-Aware Curriculum Learning (CL) framework [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Improvements of 𝐹0.5 at different difficulty intervals on NLPCC-test. G. Correlation Between the Curriculum Learning and Im￾provements Despite the fact that our methods yield improvements across all the sentence pairs, the reasons behind the improvements remain unclear. Figure2 shows the 𝐹0.5 improvements at dif￾ferent difficulty intervals on the NLPCC-test dataset. The dif￾ficulty scores are calculated by the formu… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

33 extracted references · 19 canonical work pages

  1. [10]

    Grammatical error correction via mixed-grained weighted training,

    J. Li, Q. Wang, C. Zhu, Z. Mao, and Y . Zhang, “Grammatical error correction via mixed-grained weighted training,” in Findings of the Association for Computational Linguistics: EMNLP 2023 , H. Bouamor, J. Pino, and K. Bali, Eds. Singapore: Association for Computational Linguistics, Dec. 2023, pp. 6027–6037. [Online]. Available: https://aclanthology.org/20...

    work page 2023

  2. [11]

    Data weighted training strategies for grammatical error correction,

    J. Lichtarge, C. Alberti, and S. Kumar, “Data weighted training strategies for grammatical error correction,” Transactions of the Association for Computational Linguistics , vol. 8, pp. 634–646, 2020. [Online]. Available: https://aclanthology.org/2020.tacl-1.41

    work page 2020

  3. [1]

    Correct like humans: Progressive learning framework for chinese text error correction,

    Y . Li, S. Ma, S. Chen, H. Huang, S. Huang, Y . Li, H.-T. Zheng, and Y . Shen, “Correct like humans: Progressive learning framework for chinese text error correction,” Expert Systems with Applications , p. 126039, 2024

    2024

  4. [2]

    Towards real-world writing assistance: A chinese character checking benchmark with faked and misspelled characters,

    Y . Li, Z. Xu, S. Chen, H. Huang, Y . Li, Y . Jiang, Z. Li, Q. Zhou, H.-T. Zheng, and Y . Shen, “Towards real-world writing assistance: A chinese character checking benchmark with faked and misspelled characters,” arXiv preprint arXiv:2311.11268 , 2023

    arXiv 2023

  5. [3]

    Towards End-to-End Spoken Grammatical Error Correction

    S. Bann `o, R. Ma, M. Qian, K. M. Knill, and M. J. Gales, “To- wards end-to-end spoken grammatical error correction,” arXiv preprint arXiv:2311.05550, 2023

    work page Pith review arXiv 2023

  6. [4]

    Online spelling correction for query comple- tion,

    H. Duan and B.-J. Hsu, “Online spelling correction for query comple- tion,” in Proceedings of the 20th international conference on World wide web, 2011, pp. 117–126

    work page 2011

  7. [5]

    On the (in) effectiveness of large language models for chinese text correction,

    Y . Li, H. Huang, S. Ma, Y . Jiang, Y . Li, F. Zhou, H.-T. Zheng, and Q. Zhou, “On the (in) effectiveness of large language models for chinese text correction,” arXiv preprint arXiv:2307.09007 , 2023

    arXiv 2023

  8. [6]

    The past mistake is the future wisdom: Error-driven contrastive probability optimization for chinese spell checking,

    Y . Li, Q. Zhou, Y . Li, Z. Li, R. Liu, R. Sun, Z. Wang, C. Li, Y . Cao, and H.-T. Zheng, “The past mistake is the future wisdom: Error-driven contrastive probability optimization for chinese spell checking,” arXiv preprint arXiv:2203.00991, 2022

    arXiv 2022

Show all 33 references
  1. [7]

    Cleme: debiasing multi-reference evaluation for grammatical error correction,

    J. Ye, Y . Li, Q. Zhou, Y . Li, S. Ma, H.-T. Zheng, and Y . Shen, “Cleme: debiasing multi-reference evaluation for grammatical error correction,” arXiv preprint arXiv:2305.10819 , 2023

    2023 arXiv

  2. [8]

    Linguistic rules-based corpus generation for native chinese grammatical error correction,

    S. Ma, Y . Li, R. Sun, Q. Zhou, S. Huang, D. Zhang, L. Yangning, R. Liu, Z. Li, Y . Caoet al., “Linguistic rules-based corpus generation for native chinese grammatical error correction,” arXiv preprint arXiv:2210.10442, 2022

    2022 arXiv

  3. [9]

    Learning from the dictionary: Heterogeneous knowledge guided fine-tuning for chinese spell checking,

    Y . Li, S. Ma, Q. Zhou, Z. Li, L. Yangning, S. Huang, R. Liu, C. Li, Y . Cao, and H. Zheng, “Learning from the dictionary: Heterogeneous knowledge guided fine-tuning for chinese spell checking,” arXiv preprint arXiv:2210.10320, 2022

    arXiv

  4. [12]

    Curriculum learning,

    Y . Bengio, J. Louradour, R. Collobert, and J. Weston, “Curriculum learning,” in Proceedings of the 26th annual international conference on machine learning , 2009, pp. 41–48

    2009

  5. [13]

    Exploiting curriculum learning in unsupervised neural machine translation,

    J. Lu and J. Zhang, “Exploiting curriculum learning in unsupervised neural machine translation,” in Findings of the Association for Computational Linguistics: EMNLP 2021 , M.-F. Moens, X. Huang, L. Specia, and S. W.-t. Yih, Eds. Punta Cana, Dominican Republic: Association for ...

    2021

  6. [14]

    When llms meet cunning texts: A fallacy understanding benchmark for large language models,

    Y . Li, Q. Zhou, Y . Luo, S. Ma, Y . Li, H.-T. Zheng, X. Hu, and S. Y . Philip, “When llms meet cunning texts: A fallacy understanding benchmark for large language models,” in The Thirty-eight Conference on Neural Information Processing Systems Datasets and Benchmarks Track, 2024

    2024

  7. [15]

    Active relation discovery: Towards general and label-aware open relation extraction,

    Y . Li, Y . Li, X. Chen, H.-T. Zheng, and Y . Shen, “Active relation discovery: Towards general and label-aware open relation extraction,” Knowledge-Based Systems, vol. 282, p. 111094, 2023

    2023

  8. [16]

    Ecomgpt: Instruction-tuning large language models with chain-of-task tasks for e-commerce,

    Y . Li, S. Ma, X. Wang, S. Huang, C. Jiang, H.-T. Zheng, P. Xie, F. Huang, and Y . Jiang, “Ecomgpt: Instruction-tuning large language models with chain-of-task tasks for e-commerce,” in Proceedings of the AAAI Conference on Artificial Intelligence , vol. 38, no. 17, 2024, pp. ...

    2024

  9. [17]

    Overview of the nlpcc 2018 shared task: Grammatical error correction,

    Y . Zhao, N. Jiang, W. Sun, and X. Wan, “Overview of the nlpcc 2018 shared task: Grammatical error correction,” in Natural Language Processing and Chinese Computing: 7th CCF International Conference, NLPCC 2018, Hohhot, China, August 26–30, 2018, Proceedings, Part II 7. Spring...

    2018

  10. [18]

    Mucgec: a multi-reference multi-source evaluation dataset for chinese grammatical error correction,

    Y . Zhang, Z. Li, Z. Bao, J. Li, B. Zhang, C. Li, F. Huang, and M. Zhang, “Mucgec: a multi-reference multi-source evaluation dataset for chinese grammatical error correction,” in Proceedings of the 2022 Conference of the North American Chapter of the Association for Computatio...

    2022

  11. [19]

    Dynamically composing domain- data selection with clean-data selection by “co-curricular learning

    W. Wang, I. Caswell, and C. Chelba, “Dynamically composing domain- data selection with clean-data selection by “co-curricular learning” for neural machine translation,” in Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics . Florence, Italy...

    2019

  12. [20]

    Norm-based curriculum learning for neural machine translation,

    X. Liu, H. Lai, D. F. Wong, and L. S. Chao, “Norm-based curriculum learning for neural machine translation,” in Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics , D. Jurafsky, J. Chai, N. Schluter, and J. Tetreault, Eds. Online: Associati...

    2020

  13. [21]

    Competence-based curriculum learning for neural machine translation,

    E. A. Platanios, O. Stretcu, G. Neubig, B. Poczos, and T. Mitchell, “Competence-based curriculum learning for neural machine translation,” in Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Techn...

    2019

  14. [22]

    Features and functions of the hsk dynamic composition corpus,

    B. Zhang, “Features and functions of the hsk dynamic composition corpus,” International Chinese Language Education , vol. 4, pp. 71–79, 2009

    2009

  15. [23]

    Syngec: Syntax-enhanced grammatical error correction with a tailored gec- oriented parser,

    Y . Zhang, B. Zhang, Z. Li, Z. Bao, C. Li, and M. Zhang, “Syngec: Syntax-enhanced grammatical error correction with a tailored gec- oriented parser,” in Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing , 2022, pp. 2518–2531

    2022

  16. [24]

    Heterogeneous recycle generation for Chinese grammatical error correction,

    C. Hinson, H.-H. Huang, and H.-H. Chen, “Heterogeneous recycle generation for Chinese grammatical error correction,” in Proceedings of the 28th International Conference on Computational Linguistics , D. Scott, N. Bel, and C. Zong, Eds. Barcelona, Spain (Online): International ...

    2020

  17. [25]

    Maskgec: Improving neural grammatical error correction via dynamic masking,

    Z. Zhao and H. Wang, “Maskgec: Improving neural grammatical error correction via dynamic masking,” in Proceedings of the AAAI Confer- ence on Artificial Intelligence , vol. 34, 2020, pp. 1226–1233

    2020

  18. [26]

    Sequence- to-action: Grammatical error correction with action guided sequence generation,

    J. Li, J. Guo, Y . Zhu, X. Sheng, D. Jiang, B. Ren, and L. Xu, “Sequence- to-action: Grammatical error correction with action guided sequence generation,” in Proceedings of the AAAI Conference on Artificial In- telligence, vol. 36, 2022, pp. 10 974–10 982

    2022

  19. [27]

    Position offset label prediction for grammatical error correction,

    X. Wu, J. Yu, X. Sun, and Y . Wu, “Position offset label prediction for grammatical error correction,” in Proceedings of the 29th International Conference on Computational Linguistics . International Committee on Computational Linguistics, Oct. 2022, pp. 5409–5418

    2022

  20. [28]

    From spelling to grammar: A new framework for Chinese grammatical error correction,

    X. Wu and Y . Wu, “From spelling to grammar: A new framework for Chinese grammatical error correction,” in Findings of the Association for Computational Linguistics: EMNLP 2022, Y . Goldberg, Z. Kozareva, and Y . Zhang, Eds. Abu Dhabi, United Arab Emirates: Association for Com...

    2022

  21. [29]

    TemplateGEC: Improving grammatical error correction with detection template,

    Y . Li, X. Liu, S. Wang, P. Gong, D. F. Wong, Y . Gao, H. Huang, and M. Zhang, “TemplateGEC: Improving grammatical error correction with detection template,” in Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) . As...

    2023

  22. [30]

    Training language models to follow instructions with human feedback,

    L. Ouyang, J. Wu, X. Jiang, D. Almeida, C. Wainwright, P. Mishkin, C. Zhang, S. Agarwal, K. Slama, A. Ray et al. , “Training language models to follow instructions with human feedback,”Advances in Neural Information Processing Systems , vol. 35, pp. 27 730–27 744, 2022

    2022

  23. [31]

    Gpt-4 technical report,

    J. Achiam, S. Adler, S. Agarwal, L. Ahmad, I. Akkaya, F. L. Aleman, D. Almeida, J. Altenschmidt, S. Altman, S. Anadkat et al. , “Gpt-4 technical report,” arXiv preprint arXiv:2303.08774 , 2023

    2023 arXiv

  24. [32]

    mT5: A massively multilingual pre-trained text-to-text transformer,

    L. Xue, N. Constant, A. Roberts, M. Kale, R. Al-Rfou, A. Siddhant, A. Barua, and C. Raffel, “mT5: A massively multilingual pre-trained text-to-text transformer,” in Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistic...

    2021

  25. [33]

    Cpt: A pre-trained unbalanced transformer for both chinese language understanding and generation,

    Y . Shao, Z. Geng, Y . Liu, J. Dai, F. Yang, L. Zhe, H. Bao, and X. Qiu, “Cpt: A pre-trained unbalanced transformer for both chinese language understanding and generation,” arXiv preprint arXiv:2109.05729, 2021

    2021 arXiv

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.