Unverified50% confidenceFactExact time
研究表明当虚假告警率超过30%时，团队对告警的响应速度会显著下降
1
Sources
50%
Confidence
Long-term
Relevance
8/19/2026
First Seen
Sources
Related Claims
Unverified学术研究表明，当误报率超过30%时，开发者对工具的信任度和使用意愿会显著下降68% similarUnverifiedIn a 2024 research paper, Anthropic quantified the phenomenon of model over-refusal, finding that safety-trained models had false positive refusal rates as high as 15–30% in certain scenarios64% similarUnverified斯坦福大学2019年研究发现，在实施287G协议的县域，拉丁裔居民向警察举报犯罪的概率下降约35%64% similarUnverifiedPagerDuty的数据显示高频告警团队的工程师平均任职时间比低频告警团队短40%以上63% similarUnverified若Sub-Agent的输出存在10%错误率，经过3个串行Agent后错误率可能累积至27%以上62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/772799API
curl https://kongchang.com/api/v1/knowledge/claims/772799MCP
get_claim(id=772799)