RedBench: A Universal Dataset for Comprehensive Red Teaming of Large Language Models

2026-01-07

arXiv.org

AI Summary

Existing red-teaming datasets suffer from inconsistent risk taxonomies, limited domain coverage, and outdated evaluation criteria, hindering systematic safety assessment of large language models (LLMs). To address these limitations, this work proposes RedBench—a unified red-teaming benchmark that integrates 37 authoritative data sources, comprising 29,362 attack and refusal samples. RedBench introduces, for the first time, a standardized taxonomy encompassing 22 risk categories and 19 domains. Through comprehensive data aggregation, consistent classification, and broad domain representation, RedBench significantly enhances the systematicity and comparability of LLM safety evaluations. The project publicly releases both the dataset and evaluation framework, establishing a robust baseline for developing safer and more reliable LLMs.

Technology Category

• Natural Language Processing: Safety and Robustness
• Machine Learning: Large Multimodal Models (LMMs)
• Philosophy and Ethics of AI: Safety, Robustness & Trustworthiness

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Security and Privacy: Large-scale security measurements
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

As large language models (LLMs) become integral to safety-critical applications, ensuring their robustness against adversarial prompts is paramount. However, existing red teaming datasets suffer from inconsistent risk categorizations, limited domain coverage, and outdated evaluations, hindering systematic vulnerability assessments. To address these challenges, we introduce RedBench, a universal dataset aggregating 37 benchmark datasets from leading conferences and repositories, comprising 29,362 samples across attack and refusal prompts. RedBench employs a standardized taxonomy with 22 risk categories and 19 domains, enabling consistent and comprehensive evaluations of LLM vulnerabilities. We provide a detailed analysis of existing datasets, establish baselines for modern LLMs, and open-source the dataset and evaluation code. Our contributions facilitate robust comparisons, foster future research, and promote the development of secure and reliable LLMs for real-world deployment. Code: https://github.com/knoveleng/redeval

Problem

• red teaming
• large language models
• adversarial prompts
• risk categorization
• vulnerability assessment

Innovation

• RedBench
• red teaming
• large language models
• risk taxonomy
• adversarial evaluation