Unverified80% confidenceOpinionTime unknown
当前LLM存在系统性弱点：倾向于迎合提示词中的预设前提（确认偏误），而非独立验证代码正确性
1
Sources
80%
Confidence
Long-term
Relevance
6/1/2026
First Seen
Sources
Related Entities
Related Claims
VerifiedThe fundamental reason prompt injection is difficult to defend against is that LLMs cannot reliably distinguish between data and instructions.76% similarUnverified当前LLM在代码生成上存在幻觉问题、上下文窗口限制和安全隐患三个系统性弱点74% similarUnverifiedLLMs may make incorrect judgments due to hallucinations, prompt injection, or contextual misunderstanding, and without independent verification mechanisms, faulty reasoning can directly translate into erroneous production operations72% similarUnverified在网络安全领域使用LLM作为裁判是有缺陷的，因为LLM总是会声称自己成功完成了攻击，必须为每个能力层级设计确定性的评分裁判72% similarUnverifiedKnowly faces LLM accuracy challenges, where automatic organization and comprehension may produce misleading knowledge associations or incorrect classifications.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/17966API
curl https://kongchang.com/api/v1/knowledge/claims/17966MCP
get_claim(id=17966)