Unverified50% confidenceBenchmarkExact time
当向模型提供完整公司记录时,严格准确率反而从41%下降到18%,同时召回率上升,精确率崩溃
1
Sources
50%
Confidence
Long-term
Relevance
9/25/2026
First Seen
Sources
Related Entities
Related Claims
Verified如果每步准确率为95%,经过20步循环后整体准确率会降至约36%77% similarUnverified当响应被冻结在初始策略时,增加提示广度将准确率从21.16%拉低到19.05%75% similarUnverified'99% 准确率'的模型在生产环境中仍可能频繁出错,因为剩余的 1% 往往对应现实中最关键的长尾细节场景73% similarUnverified一项针对GPT-4的研究显示,对MMLU题目进行语义等价改写后,模型准确率在部分子集上下降超过10个百分点73% similarUnverified移除候选筛选环节后,FEVER准确率下降到76.24%73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/949214API
curl https://kongchang.com/api/v1/knowledge/claims/949214MCP
get_claim(id=949214)