# ToxiCloakCN: Evaluating Robustness of Offensive Language Detection in Chinese with Cloaking Perturbations

arXiv: https://arxiv.org/abs/2406.12223 · HTML: https://mlsanity.com/paper/2406.12223/

## Key facts

- Cited by 31 papers in the map (data through 14 Oct 2026).
- Builds on 10 of its 23 references that are in the map.
- First posted on arXiv 18 Jun 2024.

## Builds on (most cited first, 10 of 10)

- [LoRA: Low-Rank Adaptation of Large Language Models](https://mlsanity.com/paper/2106.09685/) (2021)
- [Automated Hate Speech Detection and the Problem of Offensive Language](https://mlsanity.com/paper/1703.04009/) (2017)
- [BAE: BERT-based Adversarial Examples for Text Classification](https://mlsanity.com/paper/2004.01970/) (2020)
- [Racial Bias in Hate Speech and Abusive Language Detection Datasets](https://mlsanity.com/paper/1905.12516/) (2019)
- [SWSR: A Chinese Dataset and Lexicon for Online Sexism Detection](https://mlsanity.com/paper/2108.03070/) (2021)
- [DeepHate: Hate Speech Detection via Multi-Faceted Text Representations](https://mlsanity.com/paper/2103.11799/) (2021)
- [Detecting Offensive Language in Tweets Using Deep Learning](https://mlsanity.com/paper/1801.04433/) (2018)
- [Cross-Cultural Transfer Learning for Chinese Offensive Language Detection](https://mlsanity.com/paper/2303.17927/) (2023)
- [Offensive Language and Hate Speech Detection with Deep Learning and Transfer Learning](https://mlsanity.com/paper/2108.03305/) (2021)
- [Enhance Robustness of Language Models Against Variation Attack through Graph Integration](https://mlsanity.com/paper/2404.12014/) (2024)

## Built on by (most cited first, 30 of 31)

- [Exploring Multimodal Challenges in Toxic Chinese Detection: Taxonomy, Benchmark, and Findings](https://mlsanity.com/paper/2505.24341/) (2025)
- [Lost in Pronunciation: Detecting Chinese Offensive Language Disguised by Phonetic Cloaking Replacement](https://mlsanity.com/paper/2507.07640/) (2025)
- [Cross-Modal Transfer from Memes to Videos: Addressing Data Scarcity in Hateful Video Detection](https://mlsanity.com/paper/2501.15438/) (2025)
- [STATE ToxiCN: A Benchmark for Span-level Target-Aware Toxicity Extraction in Chinese Hate Speech Detection](https://mlsanity.com/paper/2501.15451/) (2025)
- [Embracing Contradiction: Theoretical Inconsistency Will Not Impede the Road of Building Responsible AI Systems](https://mlsanity.com/paper/2505.18139/) (2025)
- [Are Large Language Models Chronically Online Surfers? A Dataset for Chinese Internet Meme Explanation](https://mlsanity.com/paper/2510.00567/) (2025)
- [CSSBench: Evaluating the Safety of Lightweight LLMs against Chinese-Specific Adversarial Patterns](https://mlsanity.com/paper/2601.00588/) (2026)
- ["Newspaper Eat" Means "Not Tasty": A Taxonomy and Benchmark for Coded Language in Real-World Chinese Online Reviews](https://mlsanity.com/paper/2601.19932/) (2026)
- [Chinese Toxic Language Mitigation via Sentiment Polarity Consistent Rewrites](https://mlsanity.com/paper/2505.15297/) (2025)
- [Fine-Grained Chinese Hate Speech Understanding: Span-Level Resources, Coded Term Lexicon, and Enhanced Detection Frameworks](https://mlsanity.com/paper/2507.11292/) (2025)
- [MMBERT: Scaled Mixture-of-Experts Multimodal BERT for Robust Chinese Hate Speech Detection under Cloaking Perturbations](https://mlsanity.com/paper/2508.00760/) (2025)
- [Toxicity Red-Teaming: Benchmarking LLM Safety in Singapore's Low-Resource Languages](https://mlsanity.com/paper/2509.15260/) (2025)
- [Exploring the Capability Boundaries of LLMs in Mastering of Chinese Chouxiang Language](https://mlsanity.com/paper/2604.15841/) (2026)
- [Breaking the Cloak! Unveiling Chinese Cloaked Toxicity with Homophone Graph and Toxic Lexicon](https://mlsanity.com/paper/2505.22184/) (2025)
- [JiraiBench: A Bilingual Benchmark for Evaluating Large Language Models' Detection of Human Self-Destructive Behavior Content in Jirai Community](https://mlsanity.com/paper/2503.21679/) (2025)
- [From Ground Trust to Truth: Disparities in Offensive Language Judgments on Contemporary Korean Political Discourse](https://mlsanity.com/paper/2509.14712/) (2025)
- [Obfuscation Rules for Detecting and Detoxifying Korean Toxicity](https://mlsanity.com/paper/2510.10961/) (2025)
- [EMODIS: A Benchmark for Context-Dependent Emoji Disambiguation in Large Language Models](https://mlsanity.com/paper/2511.07193/) (2025)
- [NLP Privacy Risk Identification in Social Media (NLP-PRISM): A Survey](https://mlsanity.com/paper/2602.15866/) (2026)
- [ARGUS: Policy-Adaptive Ad Governance via Evolving Reinforcement with Adversarial Umpiring](https://mlsanity.com/paper/2605.02200/) (2026)
- [Harder to Defend: Towards Chinese Toxicity Attacks via Implicit Enhancement and Obfuscation Rewriting](https://mlsanity.com/paper/2605.22258/) (2026)
- [Beyond English and Evasion: A Human-Annotated Multi-Domain Benchmark for High-Stakes LLM Safety Evaluation in Chinese](https://mlsanity.com/paper/2605.29667/) (2026)
- [Phun-Bench: Evaluating LLMs on Phonological Understanding in Chinese](https://mlsanity.com/paper/2606.07300/) (2026)
- [Cross-Platform Chinese Offensive Comment Detection via Dual-Threshold Hard Example Mining](https://mlsanity.com/paper/2606.27629/) (2026)
- [New Terms, New Toxicity: Consensus-based Chinese Neologism Toxicity Detection via Search-Augmented LLMs](https://mlsanity.com/paper/2608.12361/) (2026)
- [SPAR-Hate: Auditor-Guided Multi-Perspective Role Reasoning for Bilingual Hate Speech Parsing](https://mlsanity.com/paper/2608.22018/) (2026)
- [EvoHarmBench: Breaking Content Moderation with Iterative Human-Like Evasion](https://mlsanity.com/paper/2608.27844/) (2026)
- [ReCAST: Restoration-aware Cascaded Stage-wise Training for Obfuscated SMS Risk Classification](https://mlsanity.com/paper/2609.04878/) (2026)
- [SinoGlyphBench: A Diagnostic Benchmark for Chinese Glyph-Level Obfuscation in Language-Model Moderation](https://mlsanity.com/paper/2609.05843/) (2026)
- [Benchmarking LLM Compliance with China AI Generated Content Regulations](https://mlsanity.com/paper/2609.19989/) (2026)

## Abstract (arXiv)

Detecting hate speech and offensive language is essential for maintaining a safe and respectful digital environment. This study examines the limitations of state-of-the-art large language models (LLMs) in identifying offensive content within systematically perturbed data, with a focus on Chinese, a language particularly susceptible to such perturbations. We introduce textsfToxiCloakCN, an enhanced dataset derived from ToxiCN, augmented with homophonic substitutions and emoji transformations, to test the robustness of LLMs against these cloaking perturbations. Our findings reveal that existing models significantly underperform in detecting offensive content when these perturbations are applied. We provide an in-depth analysis of how different types of offensive content are affected by these perturbations and explore the alignment between human and model explanations of offensiveness. Our work highlights the urgent need for more advanced techniques in offensive language detection to combat the evolving tactics used to evade detection mechanisms.