Verified65% confidenceFactExact time
ResNet在ImageNet竞赛中以3.57%的top-5错误率夺冠，首次超越人类水平（约5.1%）
3
Sources
65%
Confidence
Medium-term (~90 days)
Relevance
7/2/2026
First Seen
Valid until: 9/30/2026
Sources
PyTorch图像分类实战：数据预处理到模型训练完整流程
bilibili程序员是魔法师AI-bx6/17/2026
Related Claims
Verified2012年AlexNet在ImageNet大规模视觉识别挑战赛（ILSVRC）中以超出第二名近11个百分点的优势夺冠，将Top-5错误率从26%降至15.3%67% similarUnverified上一代Sonnet 4.5在OS World Verified基准测试中得分61.4%61% similarUnverifiedPace AI公司在保险问答基准测试中使用Sonnet 4.6达到了94%的准确率58% similarUnverified测评者在Minecraft复刻测试中给Sonnet 5评为65分（满分100），存在方块破碎无动画、无库存系统等缺陷57% similarUnverified在SWE Bench Pro编程基准测试中，Sonnet 5取得63.2%，Opus 4.8取得69.2%56% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/45211API
curl https://kongchang.com/api/v1/knowledge/claims/45211MCP
get_claim(id=45211)