2023/12/13 by Xinpeng Wang, Xiaoyuan Yi, Wang, Xinpeng +9 · 10 citations
Computer Science · Engineering · Medicine · #Artificial Intelligence (cs.AI) #Artificial intelligence #Biology #Bottleneck #Computation and Language (cs.CL) #Computer science #Context (archaeology) #Detoxification (alternative medicine) #Engineering #FOS: Computer and information sciences #Generative grammar #Hate Speech and Cyberbullying Detection #Medicine #Metric (unit) #Natural language processing #Offensive #Operations research
paper · pdf · doi:10.48550/arxiv.2312.11523
published in arXiv (Cornell University) (Cornell University)
openalex publication_date 2023/12/13 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
Warning: this paper includes model outputs showing offensive content. Recent large-scale Visual-Language Generative Models (VLGMs) have achieved unprecedented improvement in multimodal image/text generation. However, these models might also generate toxic content, e.g., offensive text and pornography images, raising significant ethical risks. Despite exhaustive studies on toxic degeneration of language models, this problem remains largely unexplored within the context of visual-language generation. This work delves into the propensity for toxicity generation and susceptibility to toxic data across various VLGMs. For this purpose, we built ToViLaG, a dataset comprising 32K co-toxic/mono-toxic text-image pairs and 1K innocuous but evocative text that tends to stimulate toxicity. Furthermore, we propose WInToRe, a novel toxicity metric tailored to visual-language generation, which theoretically reflects different aspects of toxicity considering both input and output. On such a basis, we benchmarked the toxicity of a diverse spectrum of VLGMs and discovered that some models do more evil than expected while some are more vulnerable to infection, underscoring the necessity of VLGMs detoxification. Therefore, we develop an innovative bottleneck-based detoxification method. Our method could reduce toxicity while maintaining comparable generation quality, providing a promising initial solution to this line of research.