arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2403.00829cs.AIcs.CL
收藏

TroubleLLM:对齐红队专家

TroubleLLM: Align to Red Team Expert

Ant Group(蚂蚁集团);The Chinese University of Hong Kong(香港中文大学)

Zhuoer Xu, Jianping Zhang, Shiwen Cui, Changhua Meng, Weiqiang Wang

AI总结:

针对现有LLM安全测试提示质量、多样性与可控性不足的问题,本文提出TroubleLLM,通过“用LLM测试LLM”生成可控安全测试提示,并经实验和人工评估验证其优势。

摘要:

大语言模型(LLM)已成为多种自然语言任务的最先进解决方案,并被集成到实际应用中。然而,LLM可能具有潜在危害,会表现出社会偏见和有毒内容等不良安全问题。因此,在部署前评估其安全问题势在必行。但是,现有方法生成的测试提示在质量和多样性上仍远不能令人满意。这些方法不仅耗费人力、需要大量预算成本,而且在LLM应用的特定测试领域中,测试提示生成缺乏可控性。基于“用LLM测试LLM”的思路,我们提出了首个名为TroubleLLM的LLM,用于生成有关LLM安全问题的可控测试提示。大量实验和人工评估表明,TroubleLLM在生成质量和生成可控性方面具有优越性。

英文摘要:

Large Language Models (LLMs) become the start-of-the-art solutions for a variety of natural language tasks and are integrated into real-world applications. However, LLMs can be potentially harmful in manifesting undesirable safety issues like social biases and toxic content. It is imperative to assess its safety issues before deployment. However, the quality and diversity of test prompts generated by existing methods are still far from satisfactory. Not only are these methods labor-intensive and require large budget costs, but the controllability of test prompt generation is lacking for the specific testing domain of LLM applications. With the idea of LLM for LLM testing, we propose the first LLM, called TroubleLLM, to generate controllable test prompts on LLM safety issues. Extensive experiments and human evaluation illustrate the superiority of TroubleLLM on generation quality and generation controllability.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑