Unverified50% confidenceSolutionExact time
该项目引入主动消歧机制：当不确定性超过阈值时智能体不直接输出，而是主动向用户提出澄清问题
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
8/21/2026
First Seen
Valid until: 11/19/2026
Sources
Related Claims
UnverifiedPractical questions for judging whether an AI solution is reliable include whether the team has a monitoring mechanism for hallucinations and a degradation strategy when errors occur, and the absence of these questions is itself an indicator of rhetorical packaging.77% similarUnverified通过系统提示词约束智能体只回答知识库中有的问题，可有效防止智能体回答无关问题并引导转人工76% similarUnverifiedPangu.skill introduces a 'Kaitian' mechanism, whereby when a user's question contains a false premise or is an invalid question, the system first questions and dialectically examines it rather than answering directly.73% similarUnverifiedAI对齐研究中，模型倾向回避高度不确定性的开放问题的现象被称为奖励模型过拟合或规范游戏72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/784048API
curl https://kongchang.com/api/v1/knowledge/claims/784048MCP
get_claim(id=784048)