Unverified50% confidenceFactExact time
Google的TimesFM基于PatchTST思路构建，在多个公开基准上表现优异
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/10/2026
First Seen
Valid until: 10/8/2026
Sources
Toto-2.0：多变量时间序列预测迈入基础模型规模化时代
redditr/datascience7/9/2026
Related Claims
Unverified谷歌官方披露TimesFM在200B时间点数据上进行预训练，覆盖金融、零售、能源、医疗等多个领域74% similarUnverified谷歌发布了两款基础模型TabFM（面向表格数据）和TimesFM（面向时间序列）73% similarVerified谷歌拥有自研TPU（张量处理单元）基础设施64% similarUnverified谷歌拥有自研TPU芯片集群，DeepMind提供基础研究能力，并有Search、Gmail、Google Docs等10亿级日活产品61% similarUnverified谷歌自研TPU（张量处理单元）芯片基础设施为大规模低延迟AI推理服务提供底层支撑60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/464882API
curl https://kongchang.com/api/v1/knowledge/claims/464882MCP
get_claim(id=464882)