Unverified50% confidenceBenchmarkExact time
对 177 个验证器判定结果的人工审查显示准确率达到 97.7%
1
Sources
50%
Confidence
Long-term
Relevance
10/5/2026
First Seen
Sources
Related Entities
Related Claims
Unverified证据驱动的改进前验证把精度从58.6%提升到72.3%，每轮无支持的目标从1.93个降到0.50个75% similarUnverified报告中门控准确率58%被误当成端到端质量，而真实端到端质量为98%73% similarUnverified研究表明人类评估者在判断LLM输出的事实准确性时，正确率往往只有50%-70%73% similarUnverified86%的准确率意味着仍存在14%的误差空间，Articos更适合作为早期验证和方向性判断手段，而非替代真实用户测试73% similarUnverified该模型在未见过的评估集上准确率达到96.88%73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/975471API
curl https://kongchang.com/api/v1/knowledge/claims/975471MCP
get_claim(id=975471)