[KongchangAI]
Unverified50% confidenceTradeoffExact time

当前对齐方法主要依赖训练阶段的行为塑造,在面对精心设计的对抗性提示时安全边界可能被突破,这种现象称为'越狱'(jailbreaking)

1
Sources
50%
Confidence
Long-term
Relevance
8/5/2026
First Seen

Sources

Related Claims

Cite This Claim

Stable URI
https://kongchang.com/claim/692530
API
curl https://kongchang.com/api/v1/knowledge/claims/692530
MCP
get_claim(id=692530)