Unverified50% confidenceBenchmarkExact time
在提及级别,模型整体精确率为74.1%,召回率为70.5%
1
Sources
50%
Confidence
Long-term
Relevance
9/14/2026
First Seen
Sources
Related Entities
Related Claims
Unverified在该压缩比下,模型保留了74.2%的top-1%准确率79% similarUnverified对于 7B 模型,使用求解器准确率为 75.31%,直接计算为 72.02%,配对提升 +3.29 个百分点,95% 聚类区间为 [-3.49, 10.38] 跨越零点,优势不显著77% similarUnverified一个校准良好的模型,其输出的0.8分应当在统计意义上真的对应80%的正确率73% similarUnverified当前顶级模型的单步准确率约在90%至95%之间73% similarVerified完美校准的模型在所有声称70%置信度的预测中准确率恰好为70%72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/919738API
curl https://kongchang.com/api/v1/knowledge/claims/919738MCP
get_claim(id=919738)