Unverified50% confidenceReleaseExact time
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）
1
Sources
50%
Confidence
Long-term
Relevance
7/14/2026
First Seen
Sources
Zer0Fit：用MCP将谷歌TabFM与TimesFM接入LLM聊天界面
redditr/MachineLearning7/12/2026
Related Claims
UnverifiedGoogle officially disclosed that TimesFM was pre-trained on 200B time points of data, covering multiple domains including finance, retail, energy, and healthcare.77% similarUnverifiedGoogle's TimesFM is built on the PatchTST approach and performs well on multiple public benchmarks73% similarUnverifiedGoogle Research introduced TabFM, a zero-shot tabular data foundation model employing three mechanisms—alternating row-column attention, row compression, and in-context learning Transformer—and TabFM Ensemble ranked first on both classification and regression tasks on the Tab Arena benchmark (38 classification datasets and 13 regression datasets).66% similarVerifiedThe DORA framework proposed by Google contains four key metrics: deployment frequency, change lead time, change failure rate, and service recovery time62% similarUnverifiedGoogle open-sourced the Gemma series of small models, drawing on the successful path of Meta's open release of the LLaMA series.59% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/503945API
curl https://kongchang.com/api/v1/knowledge/claims/503945MCP
get_claim(id=503945)