Estimating worst case frontier risks of open weight LLMs

【RSS 原文 · 未经 AI 筛选或翻译】In this paper, we study the worst-case frontier risks of releasing gpt-oss. We introduce malicious fine-tuning (MFT), where we attempt to elicit maximum capabilities by fine-tuning gpt-oss to be as capable as possible in two domains: biology and cybersecurity.