Unverified50% confidenceBenchmarkExact time
Training with the same recipe using only license-clean articles resulted in an accuracy drop of 0.0123 (95% interval 0.0034 to 0.0207).
1
Sources
50%
Confidence
Long-term
Relevance
10/7/2026
First Seen
Sources
Related Entities
Related Claims
Unverified在一个396个文件的大型混合职责项目上测试时，召回率几乎跌到零70% similarUnverifiedThroughout the entire training process, the flow-matching loss only slowly decreased from 0.805 to 0.754.70% similarUnverified业界报告显示即便是最先进的解析引擎，在面对非标准格式时准确率可能从95%以上降至70%以下69% similarUnverified研究表明在INT4/INT8精度下多数推理和生成任务的输出质量下降幅度在1-3%以内68% similarUnverifiedWhen using the multi-agent verification process, MARCH's accuracy was suppressed to 4.4%, whereas when the same model was directly asked which solution was better, accuracy reached 43.7%.68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/983036API
curl https://kongchang.com/api/v1/knowledge/claims/983036MCP
get_claim(id=983036)