{"url":"/task/toxic-comment-classification","name":"Toxic Comment Classification","slug":"toxic-comment-classification","description_markdown":null,"categories":[{"name":"Natural Language Processing","url":"/area/natural-language-processing"}],"source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","slug_source":"archive_url"},"counts":{"papers_tagged":28,"papers_with_code":13,"benchmarks":4,"benchmark_tables_in_archive":4,"benchmark_tables_shown":4,"benchmark_tables_withheld_as_spam":0,"benchmark_definition":"a leaderboard table with at least one row; benchmark_tables_shown also counts the zero-row tables; benchmark_tables_in_archive adds the tables withheld as spam","datasets":7,"subtasks":0,"parent_tasks":1},"benchmarks":[{"leaderboard":"/sota/toxic-comment-classification-on-civil","slug":"toxic-comment-classification-on-civil","dataset":"Civil Comments","dataset_url":"/dataset/civil-comments","rows_in_archive":22,"metrics":["GMB BPSN","GMB Subgroup","Micro F1","Precision","Recall","Macro F1","AUROC","GMB BNSP"],"first_row_in_archive_order":{"model":"RoBERTa Focal Loss","paper_title":"A benchmark for toxic comment classification on Civil Comments dataset","paper_url":"/paper/a-benchmark-for-toxic-comment-classification","paper_date":"2023-01-26","arxiv_id":"2301.11125","code_links":[{"title":"Nigiva/hatespeech-detection-models","url":"https://github.com/Nigiva/hatespeech-detection-models"}],"syntology":null}},{"leaderboard":"/sota/toxic-comment-classification-on-cad","slug":"toxic-comment-classification-on-cad","dataset":"CAD","dataset_url":"/dataset/cad","rows_in_archive":1,"metrics":["F1 on Toxic class"],"first_row_in_archive_order":{"model":"ContextRNN","paper_title":"Revisiting Contextual Toxicity Detection in Conversations","paper_url":"/paper/revisiting-contextual-toxicity-detection-in","paper_date":"2021-11-24","arxiv_id":"2111.12447","code_links":[],"syntology":null}},{"leaderboard":"/sota/toxic-comment-classification-on-germeval-2021-1","slug":"toxic-comment-classification-on-germeval-2021-1","dataset":"GermEval 2021 - Toxic Comments test set","dataset_url":null,"rows_in_archive":1,"metrics":["F1"],"first_row_in_archive_order":{"model":"GBERT/GELECTRA Ensemble","paper_title":"FHAC at GermEval 2021: Identifying German toxic, engaging, and fact-claiming comments with ensemble learning","paper_url":"/paper/fhac-at-germeval-2021-identifying-german","paper_date":"2021-09-07","arxiv_id":"2109.03094","code_links":[{"title":"dslaborg/germeval2021","url":"https://github.com/dslaborg/germeval2021"}],"syntology":null}},{"leaderboard":"/sota/toxic-comment-classification-on-jigsaw-toxic","slug":"toxic-comment-classification-on-jigsaw-toxic","dataset":"Jigsaw Toxic Comment Classification Dataset","dataset_url":"/dataset/toxic-comment-classification-challenge","rows_in_archive":1,"metrics":["Validation Accuracy"],"first_row_in_archive_order":{"model":"CapsNet","paper_title":"Evaluating The Effectiveness of Capsule Neural Network in Toxic Comment Classification using Pre-trained BERT Embeddings","paper_url":"/paper/evaluating-the-effectiveness-of-capsule","paper_date":"2023-10-12","arxiv_id":null,"code_links":[{"title":"TashinAhmed/HATE","url":"https://github.com/TashinAhmed/HATE"}],"syntology":null}}],"datasets":[{"url":"/dataset/civil-comments","name":"Civil Comments","full_name":"Jigsaw Unintended Bias in Toxicity Classification","num_papers_in_archive":156},{"url":"/dataset/cad","name":"CAD","full_name":"Contextual Abuse Dataset","num_papers_in_archive":8},{"url":"/dataset/uit-victsd","name":"UIT-ViCTSD","full_name":"UIT Vietnamese Constructive and Toxic Speech Detection","num_papers_in_archive":7},{"url":"/dataset/wikidetox","name":"WikiDetox","full_name":"Wikipedia Detox","num_papers_in_archive":7},{"url":"/dataset/toxic-comment-classification-challenge","name":"Jigsaw Toxic Comment Classification Dataset","full_name":"","num_papers_in_archive":4},{"url":"/dataset/fairprism","name":"FairPrism","full_name":"","num_papers_in_archive":2},{"url":"/dataset/korean-multi-label-hate-speech-dataset","name":"K-MHaS: Korean Multi-label Hate Speech Dataset","full_name":"","num_papers_in_archive":1}],"subtasks":[],"parent_tasks":[{"url":"/task/text-classification","name":"Text Classification"}],"papers":{"order":"repositories listed in the archive (desc), then date (desc); the archive holds no stars","population":"papers tagged with this task that list at least one repository in the archive","shown":13,"of":13,"tagged_in_all":28,"items":[{"url":"/paper/nlpguard-a-framework-for-mitigating-the-use","title":"NLPGuard: A Framework for Mitigating the Use of Protected Attributes by NLP Classifiers","date":"2024-07-01","arxiv_id":"2407.01697","repositories_listed":1,"syntology":null},{"url":"/paper/pytorch-frame-a-modular-framework-for-multi","title":"PyTorch Frame: A Modular Framework for Multi-Modal Tabular Learning","date":"2024-03-31","arxiv_id":"2404.00776","repositories_listed":1,"syntology":{"n":2,"n_ran":2,"n_unverified":0,"n_pointer_only":0}},{"url":"/paper/evaluating-the-effectiveness-of-capsule","title":"Evaluating The Effectiveness of Capsule Neural Network in Toxic Comment Classification using Pre-trained BERT Embeddings","date":"2023-10-12","arxiv_id":null,"repositories_listed":1,"syntology":null},{"url":"/paper/critic-large-language-models-can-self-correct","title":"CRITIC: Large Language Models Can Self-Correct with Tool-Interactive Critiquing","date":"2023-05-19","arxiv_id":"2305.11738","repositories_listed":1,"syntology":{"n":5,"n_ran":4,"n_unverified":1,"n_pointer_only":0}},{"url":"/paper/a-benchmark-for-toxic-comment-classification","title":"A benchmark for toxic comment classification on Civil Comments dataset","date":"2023-01-26","arxiv_id":"2301.11125","repositories_listed":1,"syntology":null},{"url":"/paper/fhac-at-germeval-2021-identifying-german","title":"FHAC at GermEval 2021: Identifying German toxic, engaging, and fact-claiming comments with ensemble learning","date":"2021-09-07","arxiv_id":"2109.03094","repositories_listed":1,"syntology":null},{"url":"/paper/data-integration-for-toxic-comment","title":"Data Integration for Toxic Comment Classification: Making More Than 40 Datasets Easily Accessible in One Unified Format","date":"2021-08-01","arxiv_id":null,"repositories_listed":1,"syntology":null},{"url":"/paper/from-hero-to-z-eroe-a-benchmark-of-low-level","title":"From Hero to Z\\'eroe: A Benchmark of Low-Level Adversarial Attacks","date":"2020-12-01","arxiv_id":null,"repositories_listed":1,"syntology":null},{"url":"/paper/from-hero-to-zeroe-a-benchmark-of-low-level","title":"From Hero to Zéroe: A Benchmark of Low-Level Adversarial Attacks","date":"2020-10-12","arxiv_id":"2010.05648","repositories_listed":1,"syntology":{"n":3,"n_ran":2,"n_unverified":1,"n_pointer_only":0}},{"url":"/paper/trojaning-language-models-for-fun-and-profit","title":"Trojaning Language Models for Fun and Profit","date":"2020-08-01","arxiv_id":"2008.00312","repositories_listed":1,"syntology":null},{"url":"/paper/can-we-achieve-more-with-less-exploring-data","title":"Can We Achieve More with Less? Exploring Data Augmentation for Toxic Comment Classification","date":"2020-07-02","arxiv_id":"2007.00875","repositories_listed":1,"syntology":null},{"url":"/paper/is-preprocessing-of-text-really-worth-your","title":"Is preprocessing of text really worth your time for online comment classification?","date":"2018-06-07","arxiv_id":"1806.02908","repositories_listed":1,"syntology":null},{"url":"/paper/convolutional-neural-networks-for-toxic","title":"Convolutional Neural Networks for Toxic Comment Classification","date":"2018-02-27","arxiv_id":"1802.09957","repositories_listed":1,"syntology":null}],"syntology_records":3,"syntology_note":"a paper without a record is not a recorded non-run: it may lack an arXiv id or simply be absent from the graph layer"},"description_links":{"kept":0,"unwrapped_to_text":0,"bare_urls_linked":0,"relative_images_dropped":0,"rule":"internal links are kept only when the target slug exists in the catalog"},"syntology":{"read_at":"2026-09-24T18:15:14+00:00","claim":"Per-sample execution status on synthesized fixtures ('ran N of M samples'); not a correctness claim and not a ranking signal.","status_vocabulary":{"ran_honours":"ran, honoured the contract we drafted","ran_violates":"ran, violated the contract we drafted","ran_draft_wrong":"ran; our contract draft was wrong, not the code","ran_fixture":"ran; our fixture could not drive it","ran":"ran on a synthesized input","unverified":"unverified (harvested, no recorded run)"}},"not_shown":{"libraries":"the archive has no per-task library table","trend_sparklines":"the Trend column of the benchmarks table was a rendered image; it is not in the archive","social_and_latest_sorts":"stars and social signals are not in the archive"}}