Unverified50% confidenceFactExact time
The current best-performing model on MNIST has an error rate below 0.2%.
1
Sources
50%
Confidence
Long-term
Relevance
8/17/2026
First Seen
Sources
Related Claims
UnverifiedIn stress testing scenarios, the error rate is typically required to be below 0.1%.67% similarUnverified在SWE-bench Full版本上，多数模型的通过率仍低于20%64% similarUnverifiedSWE-Bench Pro has a false positive rate of 8.5% and a false negative rate as high as 24%, while DeepSWE has a false positive rate of only 0.3% and a false negative rate of only 1.1%.64% similarUnverifiedZillow 的 Zestimate 自动估值模型误差率中位数在2-7%之间64% 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63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/763898API
curl https://kongchang.com/api/v1/knowledge/claims/763898MCP
get_claim(id=763898)