ToxiGAN: Toxic Data Augmentation via LLM-Guided Directional Adversarial Generation

2026-01-06

arXiv.org

AI Summary

This work addresses the limitations of existing toxic language classification methods, which suffer from a lack of class-controllable text augmentation techniques that hinder model robustness. To overcome this, the authors propose ToxiGAN, a class-aware text augmentation framework that integrates a generative adversarial network (GAN) architecture with semantic guidance from large language models (LLMs). Through a two-stage targeted adversarial training process, ToxiGAN dynamically selects LLM-generated neutral texts as semantic anchors to steer the generation of high-fidelity, class-specific toxic samples, effectively mitigating mode collapse and semantic drift. Evaluated on four hate speech benchmarks, ToxiGAN significantly outperforms current data augmentation approaches in both macro-F1 and hate-F1 metrics, demonstrating its effectiveness and novelty.

Technology Category

• Natural Language Processing: Generation
• Computer Vision: Generative Adversarial Networks (GANs) for Vision
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Search and Retrieval-Augmented AI: Large language models for search

Abstract

Augmenting toxic language data in a controllable and class-specific manner is crucial for improving robustness in toxicity classification, yet remains challenging due to limited supervision and distributional skew. We propose ToxiGAN, a class-aware text augmentation framework that combines adversarial generation with semantic guidance from large language models (LLMs). To address common issues in GAN-based augmentation such as mode collapse and semantic drift, ToxiGAN introduces a two-step directional training strategy and leverages LLM-generated neutral texts as semantic ballast. Unlike prior work that treats LLMs as static generators, our approach dynamically selects neutral exemplars to provide balanced guidance. Toxic samples are explicitly optimized to diverge from these exemplars, reinforcing class-specific contrastive signals. Experiments on four hate speech benchmarks show that ToxiGAN achieves the strongest average performance in both macro-F1 and hate-F1, consistently outperforming traditional and LLM-based augmentation methods. Ablation and sensitivity analyses further confirm the benefits of semantic ballast and directional training in enhancing classifier robustness.

Problem

• toxic data augmentation
• class-specific augmentation
• distributional skew
• limited supervision
• toxicity classification

Innovation

• ToxiGAN
• adversarial generation
• LLM-guided augmentation
• semantic ballast
• directional training