OpenAI称10月GPT-6比GPT-5.6更难越狱——但其自身测试显示部分安全性能倒退
OpenAI Says October GPT-6 Is Harder to Jailbreak Than GPT-5.6 — But Its Own Tests Show Some Safety Regressions
来源 Techgenyz
报道
- OpenAI称10月GPT-6比GPT-5.6更难越狱——但其自身测试显示部分安全性能倒退 OpenAI Says October GPT-6 Is Harder to Jailbreak Than GPT-5.6 — But Its Own Tests Show Some Safety Regressions Techgenyz · · 首发