Abstract:Inherent temporal heterogeneity, such as varying sampling densities and periodic structures, has posed substantial challenges in zero-shot generalization for Time Series Foundation Models (TSFMs). Existing TSFMs predominantly rely on massive parameterization to absorb such heterogeneity, as their static tokenization and positional encoding schemes entangle diverse temporal patterns into a fixed representation space, encouraging memorization rather than adaptation. To address this limitation, we propose Kairos, a flexible and parameter-efficient TSFM dedicated to forecasting tasks, which decouples temporal heterogeneity from model capacity through a novel tokenization perspective. Kairos introduces a dynamic patching tokenizer and a mixture-of-size encoding that adapt observational granularity to local information density, enabling fine-grained temporal abstraction without increasing model width or depth. In addition, we design a multi-granularity positional embedding based on dynamic rotary encodings, which conditions on instance-level spectral features and temporal structure induced by dynamic patching tokenization, allowing robust modeling of diverse temporal dependencies. Trained on a novel Predictability-Stratified Time-Series (PreSTS) corpus, Kairos achieves superior zero-shot performance with substantially fewer parameters on two mainstream benchmarks, GIFT-Eval and Time-Series-Library. The project page is at this https URL .
| Subjects: | Machine Learning (cs.LG) |
| Cite as: | arXiv:2509.25826 [cs.LG] |
| (or arXiv:2509.25826v3 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2509.25826 arXiv-issued DOI via DataCite |
Submission history
From: Kun Feng [view email]
[v1]
Tue, 30 Sep 2025 06:02:26 UTC (1,196 KB)
[v2]
Fri, 13 Feb 2026 16:54:37 UTC (1,120 KB)
[v3]
Thu, 14 May 2026 13:13:15 UTC (1,105 KB)