Saved in:
Bibliographic Details
Main Authors: Qiao, Zhongzheng, Liu, Chenghao, Zhang, Yiming, Jin, Ming, Pham, Quang, Wen, Qingsong, Suganthan, P. N., Jiang, Xudong, Ramasamy, Savitha
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2506.14087
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915544276402176
author Qiao, Zhongzheng
Liu, Chenghao
Zhang, Yiming
Jin, Ming
Pham, Quang
Wen, Qingsong
Suganthan, P. N.
Jiang, Xudong
Ramasamy, Savitha
author_facet Qiao, Zhongzheng
Liu, Chenghao
Zhang, Yiming
Jin, Ming
Pham, Quang
Wen, Qingsong
Suganthan, P. N.
Jiang, Xudong
Ramasamy, Savitha
contents Time series foundation models (TSFMs) demonstrate impressive zero-shot performance for time series forecasting. However, an important yet underexplored challenge is how to effectively finetune TSFMs on specific downstream tasks. While naive finetuning can yield performance gains, we argue that it falls short of fully leveraging TSFMs' capabilities, often resulting in overfitting and suboptimal performance. Given the diverse temporal patterns across sampling scales and the inherent multi-scale forecasting capabilities of TSFMs, we adopt a causal perspective to analyze finetuning process, through which we highlight the critical importance of explicitly modeling multiple scales and reveal the shortcomings of naive approaches. Focusing on encoder-based TSFMs, we propose Multiscale finetuning (MSFT), a simple yet general framework that explicitly integrates multi-scale modeling into the finetuning process. Experimental results on three different backbones (Moirai, Moment and Units) demonstrate that TSFMs finetuned with MSFT not only outperform naive and typical parameter efficient finetuning methods but also surpass state-of-the-art deep learning methods. Codes are available at https://github.com/zqiao11/MSFT.
format Preprint
id arxiv_https___arxiv_org_abs_2506_14087
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Multi-Scale Finetuning for Encoder-based Time Series Foundation Models
Qiao, Zhongzheng
Liu, Chenghao
Zhang, Yiming
Jin, Ming
Pham, Quang
Wen, Qingsong
Suganthan, P. N.
Jiang, Xudong
Ramasamy, Savitha
Machine Learning
Time series foundation models (TSFMs) demonstrate impressive zero-shot performance for time series forecasting. However, an important yet underexplored challenge is how to effectively finetune TSFMs on specific downstream tasks. While naive finetuning can yield performance gains, we argue that it falls short of fully leveraging TSFMs' capabilities, often resulting in overfitting and suboptimal performance. Given the diverse temporal patterns across sampling scales and the inherent multi-scale forecasting capabilities of TSFMs, we adopt a causal perspective to analyze finetuning process, through which we highlight the critical importance of explicitly modeling multiple scales and reveal the shortcomings of naive approaches. Focusing on encoder-based TSFMs, we propose Multiscale finetuning (MSFT), a simple yet general framework that explicitly integrates multi-scale modeling into the finetuning process. Experimental results on three different backbones (Moirai, Moment and Units) demonstrate that TSFMs finetuned with MSFT not only outperform naive and typical parameter efficient finetuning methods but also surpass state-of-the-art deep learning methods. Codes are available at https://github.com/zqiao11/MSFT.
title Multi-Scale Finetuning for Encoder-based Time Series Foundation Models
topic Machine Learning
url https://arxiv.org/abs/2506.14087