Unverified50% confidenceFactExact time
在CLINC150基准测试上，80条微调数据可使向量路由达到85.9%的准确率，完整fine-tune可达96%以上
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/2/2026
First Seen
Valid until: 9/30/2026
Sources
四层漏斗架构：生产级Agent意图路由系统设计指南
bilibili费曼学AI6/12/2026
Related Claims
Unverified测试的准确率数据在n=100样本量下误差约为±8-10个百分点（基于95%置信水平的二项分布置信区间）77% similarUnverified该系统在50个已验证的CVE样本上实现了85%以上的检测准确率76% similarUnverified在同一复现实验中，ResNet-20测试准确率为90.4%，ResNet-56测试准确率为91.7%，训练准确率分别为97.4%和99.0%75% similarUnverified研究表明经过良好校准的LLM裁判与人工标注的一致性可达80%以上74% similarUnverifiedAI意图识别准确率一般在85%-95%之间74% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/48899API
curl https://kongchang.com/api/v1/knowledge/claims/48899MCP
get_claim(id=48899)