mlsanityData to 14 Oct 2026

A Survey of Toxicity Detection and Mitigation Strategies for Multilingual Language Models

arXiv 2606.25380 ↗First posted 24 Jun 2026cs.CLcited by 0

On the map: LLM safety and security → Safety alignment, refusal and guardrails

Abstract (arXiv)

Large language models (LLMs) are increasingly deployed across languages, but their safety behavior remains uneven across linguistic and cultural contexts. This survey synthesizes work on toxicity detection and detoxification for multilingual LLMs. We first catalogue threat models that exploit language choice, translation pivots, code-switching, orthographic variation, multi-turn interaction, and post-deployment fine-tuning to weaken safety alignment. We then organize task formulations (toxic-to-neutral rewriting, toxicity classification, and toxic-generation evaluation), multilingual detection approaches (cross-lingual encoders, translation pipelines, representation-level probes, and LLM-based detectors), and mitigation strategies spanning data filtering, supervised and preference-based tuning, decoding-time steering, representation editing, and multilingual guardrails. Across these areas, we identify persistent challenges: uneven language coverage, culturally contingent definitions of harm, fragmented evaluation protocols, and the risk that detoxification suppresses legitimate dialectal or identity-related expression.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Safety alignment, refusal and guardrails, one of 12 directions in LLM safety and security. Builds on 74 of its 101 references that are in the map. First posted on arXiv 24 Jun 2026.

Builds on

74 of 101 references.

Closest ancestors · 6

Same direction, newest first

  1. 2025CREST: Universal Safety Guardrails Through Cluster-Guided Cross-Lingual Transfer2 Dec
  2. 2025MrGuard: A Multilingual Reasoning Guardrail for Universal LLM Safety21 Apr
  3. 2024Towards Understanding the Fragility of Multilingual LLMs against Fine-Tuning Attacks23 Oct
  4. 2024On Calibration of LLM-based Guard Models for Reliable Content Moderation14 Oct
  5. 2024WildGuard: Open One-Stop Moderation Tools for Safety Risks, Jailbreaks, and Refusals of LLMs26 Jun
  6. 2023Self-Detoxifying Language Models via Toxification Reversal14 Oct

Other references · 68

  1. 2022Training language models to follow instructions with human feedbackcited by 15,971
  2. 2019BERTScore: Evaluating Text Generation with BERTcited by 5,512
  3. 2019Unsupervised Cross-lingual Representation Learning at Scalecited by 3,935
  4. 2022Locating and Editing Factual Associations in GPTcited by 2,691
  5. 2020RealToxicityPrompts: Evaluating Neural Toxic Degeneration in Language Modelscited by 1,255
  6. 2023Llama Guard: LLM-based Input-Output Safeguard for Human-AI Conversationscited by 1,207
  7. 2020BLEURT: Learning Robust Metrics for Text Generationcited by 1,203
  8. 2020COMET: A Neural Framework for MT Evaluationcited by 1,033
  9. 2022Red Teaming Language Models with Language Modelscited by 1,026
  10. 2023Steering Language Models With Activation Engineeringcited by 922
  11. 2019Plug and Play Language Models: A Simple Approach to Controlled Text Generationcited by 799
  12. 2021DExperts: Decoding-Time Controlled Text Generation with Experts and Anti-Expertscited by 441
38 more
  1. 2022ToxiGen: A Large-Scale Machine-Generated Dataset for Adversarial and Implicit Hate Speech Detectioncited by 422
  2. 2023Toxicity in ChatGPT: Analyzing Persona-assigned Language Modelscited by 360
  3. 2021Annotators with Attitudes: How Annotator Beliefs And Identities Bias Toxic Language Detectioncited by 264
  4. 2021Quality at a Glance: An Audit of Web-Crawled Multilingual Datasetscited by 243
  5. 2020HateCheck: Functional Tests for Hate Speech Detection Modelscited by 216
  6. 2019SemEval-2019 Task 6: Identifying and Categorizing Offensive Language in Social Media (OffensEval)cited by 207
  7. 2024Rainbow Teaming: Open-Ended Generation of Diverse Adversarial Promptscited by 205
  8. 2022A New Generation of Perspective API: Efficient Multilingual Character-level Transformerscited by 185
  9. 2020SemEval-2020 Task 12: Multilingual Offensive Language Identification in Social Media (OffensEval 2020)cited by 138
  10. 2024AEGIS: Online Adaptive AI Content Safety Moderation with Ensemble of LLM Expertscited by 136
  11. 2021Detoxifying Language Models Risks Marginalizing Minority Voicescited by 107
  12. 2024The Language Barrier: Dissecting Safety Challenges of LLMs in Multilingual Contextscited by 94
  13. 2023Red teaming ChatGPT via Jailbreaking: Bias, Robustness, Reliability and Toxicitycited by 89
  14. 2024Detoxifying Large Language Models via Knowledge Editingcited by 87
  15. 2022Exploring the Limits of Domain-Adaptive Training for Detoxifying Large-Scale Language Modelscited by 68
  16. 2023Multilingual Large Language Models Are Not (Yet) Code-Switcherscited by 65
  17. 2025PolyGuard: A Multilingual Safety Moderation Tool for 17 Languagescited by 59
  18. 2021A Plug-and-Play Method for Controlled Text Generationcited by 56
  19. 2022Multilingual HateCheck: Functional Tests for Multilingual Hate Speech Detection Modelscited by 51
  20. 2024RLHF Can Speak Many Languages: Unlocking Multilingual Preference Optimization for LLMscited by 43
  21. 2024Code-Switching Red-Teaming: LLM Evaluation for Safety and Multilingual Understandingcited by 34
  22. 2020Fairness and Robustness in Invariant Learning: A Case Study in Toxicity Classificationcited by 32
  23. 2024Preference Tuning For Toxicity Mitigation Generalizes Across Languagescited by 31
  24. 2024Content Moderation by LLM: From Accuracy to Legitimacycited by 30
  25. 2022Toxicity in Multilingual Machine Translation at Scalecited by 26
  26. 2024"They are uncultured": Unveiling Covert Harms and Social Threats in LLM Generated Conversationscited by 25
  27. 2022APPDIA: A Discourse-aware Transformer-based Style Transfer Model for Offensive Social Media Conversationscited by 23
  28. 2023Goodtriever: Adaptive Toxicity Mitigation with Retrieval-augmented Modelscited by 23
  29. 2024Can LLMs Recognize Toxicity? A Structured Investigation Framework and Toxicity Metriccited by 23
  30. 2024Sandwich attack: Multi-language Mixture Adaptive Attack on LLMscited by 20
  31. 2024The effect of fine-tuning on language model toxicitycited by 16
  32. 2023Exploring Methods for Cross-lingual Text Style Transfer: The Case of Text Detoxificationcited by 15
  33. 2023LifeTox: Unveiling Implicit Toxicity in Life Advicecited by 14
  34. 2025Multi-lingual Multi-turn Automated Red Teaming for LLMscited by 13
  35. 2024Multilingual and Explainable Text Detoxification with Parallel Corporacited by 12
  36. 2024Text Detoxification as Style Transfer in English and Hindicited by 11
  37. 2025Steering into New Embedding Spaces: Analyzing Cross-Lingual Alignment Induced by Model Interventions in Multilingual Language Modelscited by 11
  38. 2025Breaking Bad Tokens: Detoxification of LLMs Using Sparse Autoencoderscited by 9

50 most cited of 68 All 68 (JSON)

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2024From One to Many: Expanding the Scope of Toxicity Mitigation in Language Models
  2. 2024RTP-LX: Can LLMs Evaluate Toxicity in Multilingual Scenarios?
  3. 2025Rethinking Toxicity Evaluation in Large Language Models: A Multi-Label Perspective
  4. 2024Realistic Evaluation of Toxicity in Large Language Models
  5. 2025Aligned Probing: Relating Toxic Behavior and Model Internals
  6. 2025Something Just Like TRuST : Toxicity Recognition of Span and Target
  7. 2026Beyond a Global Norm: Personalizing Toxicity Sensitivity in Language Models Without Retraining
  8. 2026HarmProfile: Characterizing Harmful Distributions in Frontier LLMs