Uncovering Zero-Shot Generalization Gaps in Time-Series Foundation Models Using Real-World Videos
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2025
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866909931924357120 |
|---|---|
| author | Li, Lujun Sleem, Lama Wang, Yiqun Xu, Yangjie Gentile, Niccolò State, Radu |
| author_facet | Li, Lujun Sleem, Lama Wang, Yiqun Xu, Yangjie Gentile, Niccolò State, Radu |
| contents | Recent research on time-series foundation models (TSFMs) has underscored the scarcity of real-world data, often supplemented with synthetic sources in existing datasets, whose generalizability remains however debated. As such, in this work, we propose a novel benchmarking approach: in particular, we aim at building a curated dataset reflecting real world physical temporal dynamics, extracting temporal signals from real-world videos using optical flow. As such, we introduce REAL-V-TSFM, a novel dataset designed to capture rich and diverse time series derived from real-world videos. Experimental results on state-of-the-art TSFMs under zero-shot forecasting show that, despite strong performance on conventional benchmarks, these models exhibit performance degradation on the proposed dataset, suggesting limited generalizability to novel datasets. These findings underscore the need for novel approaches to acquiring time series data and highlight the lack of universality in recent TSFMs, while further validating the effectiveness of our video-based time series data extraction pipeline. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2509_26347 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Uncovering Zero-Shot Generalization Gaps in Time-Series Foundation Models Using Real-World Videos Li, Lujun Sleem, Lama Wang, Yiqun Xu, Yangjie Gentile, Niccolò State, Radu Artificial Intelligence Recent research on time-series foundation models (TSFMs) has underscored the scarcity of real-world data, often supplemented with synthetic sources in existing datasets, whose generalizability remains however debated. As such, in this work, we propose a novel benchmarking approach: in particular, we aim at building a curated dataset reflecting real world physical temporal dynamics, extracting temporal signals from real-world videos using optical flow. As such, we introduce REAL-V-TSFM, a novel dataset designed to capture rich and diverse time series derived from real-world videos. Experimental results on state-of-the-art TSFMs under zero-shot forecasting show that, despite strong performance on conventional benchmarks, these models exhibit performance degradation on the proposed dataset, suggesting limited generalizability to novel datasets. These findings underscore the need for novel approaches to acquiring time series data and highlight the lack of universality in recent TSFMs, while further validating the effectiveness of our video-based time series data extraction pipeline. |
| title | Uncovering Zero-Shot Generalization Gaps in Time-Series Foundation Models Using Real-World Videos |
| topic | Artificial Intelligence |
| url | https://arxiv.org/abs/2509.26347 |