BriefGPT.xyz
May, 2024
Hacc-Man:破解LLMs的街机游戏
Hacc-Man: An Arcade Game for Jailbreaking LLMs
HTML
PDF
Matheus Valentim, Jeanette Falk, Nanna Inie
TL;DR
这篇论文介绍了一款名为 Hacc-Man 的游戏,通过挑战玩家“越狱”一个大型语言模型(LLMs),以此来提高人们对在日常系统中部署易损LLMs的风险的认识,增强人们与LLMs互动的自我效能感,并探索人们在这个新环境中采用的创造性问题解决策略。
Abstract
The recent leaps in complexity and fluency of
large language models
(LLMs) mean that, for the first time in human history, people can interact with computers using natural language alone. This creates monumental possibilities of
→