UltraSTF: Ultra -Compact Model for Large-Scale S patio- T emporal F orecasting
Chin-Chia Michael Yeh, Xiran Fan, Zhimeng Jiang, Yujie Fan, Huiyuan Chen, Uday Singh Saini, Vivian Lai, Xin Dai, Junpeng Wang, Zhongfang Zhuang, Liang Wang, Yan Zheng · 2025
Read UltraSTF: Ultra -Compact Model for Large-Scale S patio- T emporal F orecasting by Chin-Chia Michael Yeh, Xiran Fan, Zhimeng Jiang, Yujie Fan, Huiyuan Chen, Uday Singh Saini, Vivian Lai, Xin Dai, Junpeng Wang, Zhongfang Zhuang, Liang Wang, Yan Zheng (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-2502-20634.txt.
Description
Chin-Chia Michael Yeh and co-authors build UltraSTF, a tiny model for forecasting data that varies across both space and time, such as traffic, payments, and ride demand. Such data has so many channels that heavy models are wasteful, and a recent compact forecaster, SparseTSF, already saves space by leaning on repeating cycles. Its weakness is what happens inside each cycle, where it misses finer patterns.
UltraSTF adds a small bank of stored shapes that an attention step can mix and match, capturing those inner patterns while keeping the model lean. On a large benchmark it reached the best accuracy while using under 0.2 percent of the parameters of the next-best methods, stretching the trade-off between size and skill.
Catalog record
| CREATOR | Chin-Chia Michael Yeh, Xiran Fan, Zhimeng Jiang, Yujie Fan, Huiyuan Chen, Uday Singh Saini, Vivian Lai, Xin Dai, Junpeng Wang, Zhongfang Zhuang, Liang Wang, Yan Zheng |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 55 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2502-20634