HarmBench: A Standardized Evaluation Framework for Automated Red Teaming and Robust Refusal
UIUC · Center for AI Safety · CMU, Carnegie Mellon University · Tiny Fish Inc. · UC Berkeley · Microsoft · UC Berkeley, Scale AI · University of Chicago · Univeristy of Illinois at Urbana-Champaign
PDF 由论文原始站点提供，PaperCompass 不保存论文文件。
摘要
Automated red teaming holds substantial promise for uncovering and mitigating the risks associated with the malicious use of large language models (LLMs), yet the field lacks a standardized evaluation framework to rigorously assess new methods. To address this issue, we introduce HarmBench, a standardized evaluation framework for automated red teaming. We identify several desirable properties previously unaccounted for in red teaming evaluations and systematically design HarmBench to meet these criteria. Using HarmBench, we conduct a large-scale comparison of 18 red teaming methods and 33 target LLMs and defenses, yielding novel insights. We also introduce a highly efficient adversarial training method that greatly enhances LLM robustness across a wide range of attacks, demonstrating how HarmBench enables codevelopment of attacks and defenses. We open source HarmBench at https://github.com/centerforaisafety/HarmBench.