{"url":"/sota/grammatical-error-correction-on-bea-2019-test","task":{"name":"Grammatical Error Correction","url":"/task/grammatical-error-correction","note":null},"dataset":{"name":"BEA-2019 (test)","url":"/dataset/locness-corpus"},"category":"Natural Language Processing","categories":["Natural Language Processing"],"category_note":null,"description":"Grammatical Error Correction (GEC) is the task of correcting different kinds of errors in text such as spelling, punctuation, grammatical, and word choice errors. \r\n\r\nGEC is typically formulated as a sentence correction task. A GEC system takes a potentially erroneous sentence as input and is expected to transform it to its corrected version. See the example given below: \r\n\r\n| Input (Erroneous)          | Output (Corrected)     |\r\n| -------------------------  | ---------------------- |\r\n|She see Tom is catched by policeman in park at last night. | She saw Tom caught by a policeman in the park last night.|","description_from":"task","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","rank":"the archive's row order at snapshot; not re-ranked","rows_end_at":"2025-07-28","rows_withheld_as_spam":0,"metric_values":"the archive's strings, untouched"},"metrics":["F0.5"],"metric_direction":{"note":"inferred from the metric name only (the archive records no direction); null = not inferred, chart draws points only","by_metric":{"F0.5":null}},"counts":{"rows":19,"rows_with_code":18,"rows_with_paper_page":19,"rows_dated":18,"rows_using_additional_data":1},"rows":[{"rank_in_archive_order":1,"model":"Majority-voting ensemble on best 7 models","metrics":{"F0.5":"81.4"},"uses_additional_data":false,"paper_date":"2024-04-23","paper":"/paper/pillars-of-grammatical-error-correction","paper_url":"https://arxiv.org/abs/2404.14914v1","paper_title":"Pillars of Grammatical Error Correction: Comprehensive Inspection Of Contemporary Approaches In The Era of Large Language Models","code":"https://github.com/grammarly/pillars-of-gec","n_code_links":1,"syntology":null},{"rank_in_archive_order":2,"model":"GRECO (voting+ESC)","metrics":{"F0.5":"80.84"},"uses_additional_data":false,"paper_date":"2023-10-23","paper":"/paper/system-combination-via-quality-estimation-for","paper_url":"https://arxiv.org/abs/2310.14947v1","paper_title":"System Combination via Quality Estimation for Grammatical Error Correction","code":"https://github.com/nusnlp/greco","n_code_links":1,"syntology":null},{"rank_in_archive_order":3,"model":"ESC","metrics":{"F0.5":"79.90"},"uses_additional_data":false,"paper_date":null,"paper":"/paper/frustratingly-easy-system-combination-for","paper_url":"https://aclanthology.org/2022.naacl-main.143","paper_title":"Frustratingly Easy System Combination for Grammatical Error Correction","code":"https://github.com/nusnlp/esc","n_code_links":1,"syntology":null},{"rank_in_archive_order":4,"model":"RedPenNet","metrics":{"F0.5":"77.60"},"uses_additional_data":true,"paper_date":"2023-09-19","paper":"/paper/redpennet-for-grammatical-error-correction","paper_url":"https://arxiv.org/abs/2309.10898v1","paper_title":"RedPenNet for Grammatical Error Correction: Outputs to Tokens, Attentions to Spans","code":"https://github.com/webspellchecker/unlp-2023-shared-task","n_code_links":1,"syntology":null},{"rank_in_archive_order":5,"model":"clang_large_ft2-gector","metrics":{"F0.5":"77.1"},"uses_additional_data":false,"paper_date":"2021-11-16","paper":"/paper/improved-grammatical-error-correction-by","paper_url":"https://openreview.net/forum?id=bg470UXkLqF","paper_title":"Improved grammatical error correction by ranking elementary edits","code":"https://github.com/AlexeySorokin/EditScorer","n_code_links":1,"syntology":null},{"rank_in_archive_order":6,"model":"Unsupervised GEC + cLang8","metrics":{"F0.5":"76.5"},"uses_additional_data":false,"paper_date":"2023-12-06","paper":"/paper/unsupervised-grammatical-error-correction","paper_url":"https://aclanthology.org/2023.emnlp-main.185.pdf","paper_title":"Unsupervised Grammatical Error Correction Rivaling Supervised Methods","code":"https://github.com/nusnlp/ugec","n_code_links":1,"syntology":null},{"rank_in_archive_order":7,"model":"DeBERTa + RoBERTa + XLNet","metrics":{"F0.5":"76.05"},"uses_additional_data":false,"paper_date":"2022-03-24","paper":"/paper/ensembling-and-knowledge-distilling-of-large-1","paper_url":"https://arxiv.org/abs/2203.13064v1","paper_title":"Ensembling and Knowledge Distilling of Large Sequence Taggers for Grammatical Error Correction","code":"https://github.com/makstarnavskyi/gector-large","n_code_links":1,"syntology":null},{"rank_in_archive_order":8,"model":"MoECE","metrics":{"F0.5":"74.07"},"uses_additional_data":false,"paper_date":"2024-10-30","paper":"/paper/efficient-and-interpretable-grammatical-error","paper_url":"https://arxiv.org/abs/2410.23507v1","paper_title":"Efficient and Interpretable Grammatical Error Correction with Mixture of Experts","code":"https://github.com/nusnlp/moece","n_code_links":1,"syntology":null},{"rank_in_archive_order":9,"model":"Sequence tagging + token-level transformations + two-stage fine-tuning (+RoBERTa, XLNet)","metrics":{"F0.5":"73.7"},"uses_additional_data":false,"paper_date":"2020-05-26","paper":"/paper/gector-grammatical-error-correction-tag-not","paper_url":"https://arxiv.org/abs/2005.12592v2","paper_title":"GECToR -- Grammatical Error Correction: Tag, Not Rewrite","code":"https://github.com/grammarly/gector","n_code_links":3,"syntology":{"n_ran":2,"n_unverified":10,"n_samples":12,"n_pointer_only_licence":0}},{"rank_in_archive_order":10,"model":"BEA Combination","metrics":{"F0.5":"73.2"},"uses_additional_data":false,"paper_date":"2019-06-10","paper":"/paper/learning-to-combine-grammatical-error","paper_url":"https://arxiv.org/abs/1906.03897v1","paper_title":"Learning to combine Grammatical Error Corrections","code":"https://github.com/IBM/learning-to-combine-grammatical-error-corrections","n_code_links":1,"syntology":null},{"rank_in_archive_order":11,"model":"GEC-DI (LM+GED)","metrics":{"F0.5":"73.1"},"uses_additional_data":false,"paper_date":"2023-10-23","paper":"/paper/improving-seq2seq-grammatical-error","paper_url":"https://arxiv.org/abs/2310.14534v1","paper_title":"Improving Seq2Seq Grammatical Error Correction via Decoding Interventions","code":"https://github.com/Jacob-Zhou/gecdi","n_code_links":1,"syntology":null},{"rank_in_archive_order":12,"model":"LM-Critic","metrics":{"F0.5":"72.9"},"uses_additional_data":false,"paper_date":"2021-09-14","paper":"/paper/lm-critic-language-models-for-unsupervised","paper_url":"https://arxiv.org/abs/2109.06822v2","paper_title":"LM-Critic: Language Models for Unsupervised Grammatical Error Correction","code":"https://github.com/grammarly/gector","n_code_links":3,"syntology":null},{"rank_in_archive_order":13,"model":"Sequence tagging + token-level transformations + two-stage fine-tuning (+XLNet)","metrics":{"F0.5":"72.4"},"uses_additional_data":false,"paper_date":"2020-05-26","paper":"/paper/gector-grammatical-error-correction-tag-not","paper_url":"https://arxiv.org/abs/2005.12592v2","paper_title":"GECToR -- Grammatical Error Correction: Tag, Not Rewrite","code":"https://github.com/grammarly/gector","n_code_links":3,"syntology":{"n_ran":2,"n_unverified":10,"n_samples":12,"n_pointer_only_licence":0}},{"rank_in_archive_order":14,"model":"Transformer + Pre-train with Pseudo Data","metrics":{"F0.5":"70.2"},"uses_additional_data":false,"paper_date":"2019-09-02","paper":"/paper/an-empirical-study-of-incorporating-pseudo","paper_url":"https://arxiv.org/abs/1909.00502v1","paper_title":"An Empirical Study of Incorporating Pseudo Data into Grammatical Error Correction","code":"https://github.com/butsugiri/gec-pseudodata","n_code_links":1,"syntology":{"n_ran":3,"n_unverified":0,"n_samples":3,"n_pointer_only_licence":0}},{"rank_in_archive_order":15,"model":"Transformer + Pre-train with Pseudo Data (+BERT)","metrics":{"F0.5":"69.8"},"uses_additional_data":false,"paper_date":"2020-05-03","paper":"/paper/encoder-decoder-models-can-benefit-from-pre","paper_url":"https://arxiv.org/abs/2005.00987v2","paper_title":"Encoder-Decoder Models Can Benefit from Pre-trained Masked Language Models in Grammatical Error Correction","code":"https://github.com/kanekomasahiro/bert-gec","n_code_links":1,"syntology":null},{"rank_in_archive_order":16,"model":"Transformer","metrics":{"F0.5":"69.5"},"uses_additional_data":false,"paper_date":"2019-08-01","paper":"/paper/neural-grammatical-error-correction-systems","paper_url":"https://aclanthology.org/W19-4427","paper_title":"Neural Grammatical Error Correction Systems with Unsupervised Pre-training on Synthetic Data","code":"https://github.com/grammatical/pretraining-bea2019","n_code_links":1,"syntology":null},{"rank_in_archive_order":17,"model":"Transformer","metrics":{"F0.5":"69.0"},"uses_additional_data":false,"paper_date":"2019-07-02","paper":"/paper/a-neural-grammatical-error-correction-system","paper_url":"https://arxiv.org/abs/1907.01256v1","paper_title":"A Neural Grammatical Error Correction System Built On Better Pre-training and Sequential Transfer Learning","code":"https://github.com/kakaobrain/helo-word","n_code_links":2,"syntology":null},{"rank_in_archive_order":18,"model":"VERNet","metrics":{"F0.5":"68.9"},"uses_additional_data":false,"paper_date":"2021-05-10","paper":"/paper/neural-quality-estimation-with-multiple","paper_url":"https://arxiv.org/abs/2105.04443v1","paper_title":"Neural Quality Estimation with Multiple Hypotheses for Grammatical Error Correction","code":"https://github.com/thunlp/VERNet","n_code_links":1,"syntology":null},{"rank_in_archive_order":19,"model":"Ensemble of models","metrics":{"F0.5":"66.78"},"uses_additional_data":false,"paper_date":"2019-08-01","paper":"/paper/the-laix-systems-in-the-bea-2019-gec-shared","paper_url":"https://aclanthology.org/W19-4416","paper_title":"The LAIX Systems in the BEA-2019 GEC Shared Task","code":null,"n_code_links":0,"syntology":null}],"since_archive":{"claim":"Results that newer papers report for their own method, placed here by Syntology. A model pointed at the cell in the paper's own table; the number was read from that cell and checked against this leaderboard's metric, dataset, split and scale; an independent check that saw this leaderboard's other rows and every other leaderboard on the same dataset accepted it. Not reviewed by the paper's authors or by the archive's editors, and not ranked against the archive rows.","extraction_file_present":true,"measurement":{"test_papers":883,"papers_with_output":881,"judged_true":108,"judged":110,"wilson95_lower":0.9361,"measured_on":"2026-09-24","frozen_commit":"0e3de0df94"},"measurement_note":"blind adjudication of accepted entries on a held-out split of archive papers, rules frozen before the test","coverage":{"sentence":"Syntology has checked 6,885 of the 9,623 papers on this site that are newer than the archive; results from the others appear after they are checked.","complete":false,"papers_newer_than_archive":9623,"papers_checked":6885,"papers_extracted_not_yet_verified":0,"boards_without_verdict":2,"papers_not_yet_extracted":2737},"order":"newest first by month (arXiv date, else the arXiv-id month), then arXiv id descending","columns":[],"entries":[]},"syntology":{"read_at":"2026-09-25T09:33:49+00:00","claim":"Per row: N of M harvested code samples from that row's paper executed on a synthesized fixture; the other M-N are unverified. Not a reproduction of the row's number; not a correctness claim. n_pointer_only_licence counts samples the site points at rather than redistributes (a licence axis, independent of ran/unverified).","rows_with_graph_line":3,"rows_with_any_sample_ran":3,"distinct_papers_with_graph_line":2,"distinct_papers_with_any_sample_ran":2,"samples_over_distinct_papers":{"n_ran":5,"n_unverified":10,"n_samples":15,"n_pointer_only_licence":0,"note":"each paper (arXiv id) counted once, however many rows it is behind; this is the page-level figure"},"samples_row_weighted":{"n_ran":7,"n_unverified":20,"n_samples":27,"n_pointer_only_licence":0,"note":"row-weighted: a paper behind several rows is counted once per row; inflated relative to samples_over_distinct_papers by design, kept for readers summing the per-row syntology blocks"}}}