This Time is Different: An Observability Perspective on Time Series Foundation Models
Ben Cohen, Emaad Khwaja, Salahidine Lemaachi, Chris Lettieri, Charles Masson, Elise Ramé, Qiqi Ren, Afshin Rostamizadeh, Anna-Monica Toon, Kan Wang, Stephan Xie, Zongzhe Xu, Ameet Talwalkar, Othmane Abou-Amal · 2025
Read This Time is Different: An Observability Perspective on Time Series Foundation Models by Ben Cohen, Emaad Khwaja, Salahidine Lemaachi, Chris Lettieri, Charles Masson, Elise Ramé, Qiqi Ren, Afshin Rostamizadeh, Anna-Monica Toon, Kan Wang, Stephan Xie, Zongzhe Xu, Ameet Talwalkar, Othmane Abou-Amal (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2505-14766.txt.
Description
Ben Cohen and a large team of co-authors present a 2025 paper on Toto, a foundation model for forecasting time series from system telemetry. Monitoring data are spiky, multi channel, and huge. The model uses a modern decoder design tuned for those quirks.
Its training mix blends telemetry, open sets, and made up data, and runs several times larger than rival corpora. The team also release BOOM, a benchmark of hundreds of millions of real readings. Toto tops that benchmark and the standard public ones. Weights, code, and data are all open. The authors frame observability data as a first class training source, not scraped leftovers.
Catalog record
| CREATOR | Ben Cohen, Emaad Khwaja, Salahidine Lemaachi, Chris Lettieri, Charles Masson, Elise Ramé, Qiqi Ren, Afshin Rostamizadeh, Anna-Monica Toon, Kan Wang, Stephan Xie, Zongzhe Xu, Ameet Talwalkar, Othmane Abou-Amal |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 104 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2505-14766