Unverified50% confidenceBenchmarkExact time
Precision Mode 在一组涵盖约9,000份复杂文档的基准测试中达到了94.7%的提取准确率
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/12/2026
First Seen
Valid until: 12/11/2026
Sources
Related Entities
Related Claims
UnverifiedPrecision Mode 相比最强的前沿模型「分块-合并」（chunk-and-merge）基线方案，准确率领先7个百分点72% similarUnverified4Bit量化下模型的基准测试得分通常能保持原始精度的90%-95%70% similarUnverifiedEmsyAI使用FP16混合精度训练了15000步，验证困惑度收敛到5.8667% similarUnverifiedtop-1%准确率是评估量化模型与原始全精度模型输出一致性的指标，关注模型置信度最高的1%的token是否与全精度版本一致66% similarUnverified开发者宣称MiAPI在SimpleQA基准测试上达到82%的准确率并超越Perplexity66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/915616API
curl https://kongchang.com/api/v1/knowledge/claims/915616MCP
get_claim(id=915616)