Verified90% confidenceFactExact time
AlphaFold2在CASP14竞赛中以压倒性优势超越所有传统方法，预测精度接近实验测量水平
6
Sources
90%
Confidence
Long-term
Relevance
5/22/2026
First Seen
Sources
你最希望AI解决什么问题？一场关于AI未来方向的深度思考
twittersama5/22/2026
Related Entities
Related Claims
Unverified引导模式下平均端到端准确率从28.6%提升到33.7%，歧义检测F1值从45.3%提升到64.9%57% similarUnverifiedCAPI-DINO的ViT-B/14模型在iNat21数据集上取得了68.7%的线性探测准确率56% similarUnverifiedLFM2.5在多项基准测试中展现出与体量高达其4倍模型相当的竞争力56% similarVerifiedCodex在HumanEval基准测试上达到了28.8%的pass@1准确率56% similarUnverifiedFP8量化相比FP16原始精度，在主流benchmark上的性能下降通常不超过1-2%55% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/463API
curl https://kongchang.com/api/v1/knowledge/claims/463MCP
get_claim(id=463)