TensorX
返回文献探索

Paper · arXiv 2311.07689

MART: Improving LLM Safety with Multi-round Automatic Red-Teaming

Suyu Ge, Chunting Zhou, Rui Hou, Madian Khabsa, Yi-Chia Wang, Qifan Wang, Jiawei Han, Yuning Mao

9 upvotesNovember 13, 2023arXiv 预印本
AI 摘要

A multi-round automatic red-teaming method uses iterative adversarial prompts and safety fine-tuning to enhance the safety of large language models with minimal impact on their helpfulness.

Large Language Models (LLMs)automatic red-teamingadversarial prompt writingsafe response generationadversarial LLMsafety alignmentadversarial prompt benchmarks

Abstract

Red-teaming is a common practice for mitigating unsafe behaviors in Large Language Models (LLMs), which involves thoroughly assessing LLMs to identify potential flaws and addressing them with responsible and accurate responses. While effective, manual red-teaming is costly, and existing automatic red-teaming typically discovers safety risks without addressing them. In this paper, we propose a Multi-round Automatic Red-Teaming (MART) method, which incorporates both automatic adversarial prompt writing and safe response generation, significantly increasing red-teaming scalability and the safety of the target LLM. Specifically, an adversarial LLM and a target LLM interplay with each other in an iterative manner, where the adversarial LLM aims to generate challenging prompts that elicit unsafe responses from the target LLM, while the target LLM is fine-tuned with safety aligned data on these adversarial prompts. In each round, the adversarial LLM crafts better attacks on the updated target LLM, while the target LLM also improves itself through safety fine-tuning. On adversarial prompt benchmarks, the violation rate of an LLM with limited safety alignment reduces up to 84.7% after 4 rounds of MART, achieving comparable performance to LLMs with extensive adversarial prompt writing. Notably, model helpfulness on non-adversarial prompts remains stable throughout iterations, indicating the target LLM maintains strong performance on instruction following.

北京市昌平区探索星信息技术及软件开发工作室

京ICP备2026059466号
MART: Improving LLM Safety with Multi-round Automatic Red-Teaming | TensorX