待验证50% 置信事实精确时间
AI声称'代码已完成'但未实际验证的现象在学术界被称为'幻觉式完成'(Completion Hallucination),是大语言模型在代码生成场景中最危险的行为模式之一
1
来源数
50%
置信度
中期 (~90 天)
时效性
2026/7/2
首次发现
有效期至:2026/9/30
来源
Superpowers:给AI编程助手装上工作规范
bilibili飞猿学AI2026/6/12
相关事实
待验证AI-generated content is prone to hallucinations, confidently outputting information that seems plausible but is factually incorrect74% 相似待验证AI coding tools such as Cursor and VS Code are commonly affected by hallucination problems including fabricated API endpoints and outdated library functions74% 相似已验证AI幻觉(Hallucination)是AI编程问题的核心根源之一,表现为调用不存在的API、引用已废弃的库版本、或构造逻辑上自洽但业务上错误的算法73% 相似已验证AI代码生成的幻觉率与上下文的歧义程度正相关,项目约定越模糊模型越倾向于发明看似合理的实现73% 相似
引用此条事实
Stable URI
https://kongchang.com/claim/48049API
curl https://kongchang.com/api/v1/knowledge/claims/48049MCP
get_claim(id=48049)