mlsanityData to 7 Oct 2026

LLM in the Loop: Creating the ParaDeHate Dataset for Hate Speech Detoxification

arXiv 2506.01484 ↗First posted 2 Jun 2025cs.CLcited by 4

Not on the map. Why?

Abstract (arXiv)

Detoxification, the task of rewriting harmful language into non-toxic text, has become increasingly important amid the growing prevalence of toxic content online. However, high-quality parallel datasets for detoxification, especially for hate speech, remain scarce due to the cost and sensitivity of human annotation. In this paper, we propose a novel LLM-in-the-loop pipeline leveraging GPT-4o-mini for automated detoxification. We first replicate the ParaDetox pipeline by replacing human annotators with an LLM and show that the LLM performs comparably to human annotation. Building on this, we construct ParaDeHate, a large-scale parallel dataset specifically for hatespeech detoxification. We release ParaDeHate as a benchmark of over 8K hate/non-hate text pairs and evaluate a wide range of baseline methods. Experimental results show that models such as BART, fine-tuned on ParaDeHate, achieve better performance in style accuracy, content preservation, and fluency, demonstrating the effectiveness of LLM-generated detoxification text as a scalable alternative to human annotation.

Key facts. Cited by 4 papers in the map (data through 7 Oct 2026). Builds on 45 of its 61 references that are in the map. First posted on arXiv 2 Jun 2025.

Builds on

45 of 61 references.

References · 45

  1. 2019Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformercited by 14,491
  2. 2019RoBERTa: A Robustly Optimized BERT Pretraining Approachcited by 13,266
  3. 2019Sentence-BERT: Sentence Embeddings using Siamese BERT-Networkscited by 9,049
  4. 2023Judging LLM-as-a-Judge with MT-Bench and Chatbot Arenacited by 8,393
  5. 2019BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Generation, Translation, and Comprehensioncited by 5,874
  6. 2018Neural Network Acceptability Judgmentscited by 952
  7. 2023ChatGPT Outperforms Crowd-Workers for Text-Annotation Taskscited by 849
  8. 2017Automated Hate Speech Detection and the Problem of Offensive Languagecited by 724
  9. 2020Language-agnostic BERT Sentence Embeddingcited by 621
  10. 2024From Generation to Judgment: Opportunities and Challenges of LLM-as-a-judgecited by 480
  11. 2022ToxiGen: A Large-Scale Machine-Generated Dataset for Adversarial and Implicit Hate Speech Detectioncited by 419
  12. 2020HateXplain: A Benchmark Dataset for Explainable Hate Speech Detectioncited by 403
33 more
  1. 2024MLLM-as-a-Judge: Assessing Multimodal LLM-as-a-Judge with Vision-Language Benchmarkcited by 362
  2. 2024Humans or LLMs as the Judge? A Study on Judgement Biasescited by 314
  3. 2018Large Scale Crowdsourcing and Characterization of Twitter Abusive Behaviorcited by 265
  4. 2017Style Transfer from Non-Parallel Text by Cross-Alignmentcited by 246
  5. 2020HateCheck: Functional Tests for Hate Speech Detection Modelscited by 215
  6. 2018Dear Sir or Madam, May I introduce the GYAFC Dataset: Corpus, Benchmarks and Metrics for Formality Style Transfercited by 209
  7. 2019SemEval-2019 Task 6: Identifying and Categorizing Offensive Language in Social Media (OffensEval)cited by 206
  8. 2024Meta-Rewarding Language Models: Self-Improving Alignment with LLM-as-a-Meta-Judgecited by 202
  9. 2024Large Language Models for Data Annotation and Synthesis: A Surveycited by 201
  10. 2018Delete, Retrieve, Generate: A Simple Approach to Sentiment and Style Transfercited by 191
  11. 2020Deep Learning for Text Style Transfer: A Surveycited by 161
  12. 2017Style Transfer in Text: Exploration and Evaluationcited by 145
  13. 2019Automatically Neutralizing Subjective Bias in Textcited by 105
  14. 2023Revisiting Relation Extraction in the era of Large Language Modelscited by 100
  15. 2023LLMaAA: Making Large Language Models as Active Annotatorscited by 78
  16. 2020A Probabilistic Formulation of Unsupervised Text Style Transfercited by 70
  17. 2019A Dual Reinforcement Learning Framework for Unsupervised Text Style Transfercited by 65
  18. 2018Fighting Offensive Language on Social Media with Unsupervised Text Style Transfercited by 55
  19. 2021Text Detoxification using Large Pre-trained Neural Modelscited by 52
  20. 2021XFORMAL: A Benchmark for Multilingual Formality Style Transfercited by 51
  21. 2019"Mask and Infill" : Applying Masked Language Model to Sentiment Transfercited by 48
  22. 2023LongForm: Effective Instruction Tuning with Reverse Instructionscited by 47
  23. 2023Exploring Cross-Cultural Differences in English Hate Speech Annotations: From Dataset Construction to Analysiscited by 44
  24. 2020Parallel Data Augmentation for Formality Style Transfercited by 39
  25. 2022Large-Scale Hate Speech Detection with Cross-Domain Transfercited by 39
  26. 2021Civil Rephrases Of Toxic Texts With Self-Supervised Transformerscited by 32
  27. 2023Making Large Language Models Better Data Creatorscited by 28
  28. 2024Style-Specific Neurons for Steering LLMs in Text Style Transfercited by 24
  29. 2017Evaluating prose style transfer with the Biblecited by 23
  30. 2023A Ship of Theseus: Curious Cases of Paraphrasing in LLM-Generated Textscited by 23
  31. 2020Towards A Friendly Online Community: An Unsupervised Style Transfer Framework for Profanity Redactioncited by 17
  32. 2024MultiParaDetox: Extending Text Detoxification with Parallel Data to New Languagescited by 13
  33. 2024BMIKE-53: Investigating Cross-Lingual Knowledge Editing with In-Context Learningcited by 13

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 2; 2026: 2.

4 papers.

  1. 2025CoDAE: Adapting Large Language Models for Education via Chain-of-Thought Data Augmentationcited by 4
  2. 2025Beyond Over-Refusal: Scenario-Based Diagnostics and Post-Hoc Mitigation for Exaggerated Refusals in LLMscited by 3
  3. 2026Analyzing Bias in False Refusal Behavior of Large Language Models for Hate Speech Detoxificationcited by 1
  4. 2026Bias Dynamics in BabyLMs: Towards a Compute-Efficient Sandbox for Democratising Pre-Training Debiasingcited by 0