aiqbench analysis
This research introduces SIFT, a fine-tuning technique designed to prevent Time Series Foundation Models from overfitting or collapsing into mean predictions during adaptation. By leveraging semantic spectrum decomposition and component-wise mixup, the method effectively preserves structural fidelity while enhancing robustness against perturbations in non-core subspaces. The approach demonstrates significant performance gains across diverse real-world datasets, addressing critical limitations in current naive fine-tuning strategies for temporal data.
SIFT: Enhancing Time Series Foundation Models via Semantic Invariance and Structural Fidelity Fine-Tuning
arXiv:2609.32676v1 Announce Type: new Abstract: Time Series Foundation Models (TSFMs) have achieved remarkable zero-shot performance through extensive pre-training on massive time series datasets. Nevertheless, due to the low-dimensional properties and diverse structural patterns of time series data, performing naive fine-tuning on TSFMs often leads to overfitting and falling into the mean-prediction trap. To address these challenges, we propose SIFT, a robust adaptation method that enhances time series foundation models by preserving Semantic Invariance and structural Fidelity throughout the fine-Tuning process. We employ semantic-invariant adversarial augmentation, which utilizes semantic spectrum decomposition to partition the semantic space and then generates perturbations within the non-core semantic subspace to bolster the model's robustness against these perturbations, mitigating overfitting. We implement a component-based structural fidelity enhancement, which facilitates component-wise mixup and imposes a reconstruction objective to improve the model's ability to preserve structural fidelity, alleviating the mean-prediction trap. Extensive experiments on representative TSFMs covering 10 real-world datasets demonstrate that SIFT can significantly enhance the performance of TSFMs.
Abstract:Time Series Foundation Models (TSFMs) have achieved remarkable zero-shot performance through extensive pre-training on massive time series datasets. Nevertheless, due to the low-dimensional properties and diverse structural patterns of time series data, performing naive fine-tuning on TSFMs often leads to overfitting and falling into the mean-prediction trap. To address these challenges, we propose SIFT, a robust adaptation method that enhances time series foundation models by preserving Semantic Invariance and structural Fidelity throughout the fine-Tuning process. We employ semantic-invariant adversarial augmentation, which utilizes semantic spectrum decomposition to partition the semantic space and then generates perturbations within the non-core semantic subspace to bolster the model's robustness against these perturbations, mitigating overfitting. We implement a component-based structural fidelity enhancement, which facilitates component-wise mixup and imposes a reconstruction objective to improve the model's ability to preserve structural fidelity, alleviating the mean-prediction trap. Extensive experiments on representative TSFMs covering 10 real-world datasets demonstrate that SIFT can significantly enhance the performance of TSFMs.
Machine Learning (cs.LG)
arXiv:2609.32676 [cs.LG]
(or arXiv:2609.32676v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2609.32676 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 251 words
Related coverage
News
Quantum Computing Report · Jul 25, 2026
KairosAgent: Agentic Time Series Forecasting with Fused Semantic Reasoning
arXiv · Sep 11, 2026
Foundation Models and Fine-Tuning: Toward a New Generation of Models for Time Series Forecasting
arXiv · Jul 28, 2026
Real-Time Intelligence with IBM Time Series Models on Confluent
Hugging Face · Sep 2, 2026