SWIFT: Sample-Wise Instance Fine-Tuning of Time-Series Foundation Models for Forecasting
Authors: Yang Chen, Feifan Tian, Cong Wang, Cheng Feng, Wenbin Li, Yang Gao
Organizations: State Key Laboratory for Novel Software Technology, Nanjing University, Nanjing, 210008, China · School of Intelligent Science and Technology, Nanjing University, Suzhou, 215163, China · Siemens Data and AI Research, Beijing, 100102, China · Nanjing University– Siemens Joint Research Center on Industrial AI, Suzhou, 215163, China
While current few-shot adaptation strategies for time series forecasting foundation models are predominantly dataset-centric, the necessity of obtaining an additional dataset for the finetuning process poses a significant hurdle for adoption. To address this constraint, we propose Sample-Wise Instance Fine-Tuning (SWIFT), a novel forecasting adaptation framework that achieves fine-grained specialization by leveraging structural priors from the target instance's own history. This shift allows the fine-tuning process to be entirely transparent to the end-user, which significantly enhances the ease of adoption. To implement this framework, SWIFT synergizes Self-Retrieval for historical mining and Frequency-Aware Loss for spectral consistency, while employing Deviation-Aware Dynamic Fusion to balance adaptation with general knowledge preservation. Extensive experiments across four TSFM backbones and three long-history benchmark subsets validate SWIFT, yielding average MASE and WQL reductions of 6.41% and 4.83% across 24 combinations of four backbones and six Chronos Benchmark II datasets. To the best of our knowledge, SWIFT is the first framework to investigate sample-wise instance fine-tuning of time-series foundation models, specializing the model for an individual prediction window using only the corresponding forecasting sample's own observed history.
Figures & tables
Figure 1: Learning from others versus learning from self. The illustrated support-set adaptation learns from a separately supplied support set Dsupport of similar sequences, leveraging cross-instance information. In contrast, SWIFT constructs an intrinsic support set Dintr by mining context–target pairs from the target instance’s own history. Both approaches update only LoRA parameters, while SWIFT constructs its support set from the current sample’s own history to specialize the model for its prediction window.
Adaptation Strategy
Update Stage
External Data
Adaptation Data
Supervision Type
Adaptation Granularity
Zero-shot
None
None
None
None
N/A
Few-shot adaptation
Offline
Typically required
Support Set
Observed targets
Dataset-level
TTA
Online
None
Test Stream
Self-supervision
Stream/batch-level
SWIFT (Ours)
Pre-inference
None
Target History
Internal Priors
Instance-level
Table I: Comparison of SWIFT with representative adaptation paradigms for TSFMs.
Figure 2: Overview of SWIFT. The framework consists of three stages: (a) Self-Retrieval constructs an intrinsic support set from similar historical and recent subsequences; (b) Frequency-Aware Loss adapts the model with FreAL to emphasize informative spectral discrepancies, together with an anchor regularizer that limits prediction drift; and (c) Deviation-Aware Dynamic Fusion combines adapted and zero-shot predictions to reduce harmful over-adaptation.
Method
ERCOT
Electricity
Exchange
ETTm
ETTh
Traffic
MASE
WQL
MASE
WQL
MASE
WQL
MASE
WQL
MASE
WQL
MASE
WQL
Chronos-2
0.6275
0.0307
1.8293
0.0117
0.5813
0.0456
0.7599
0.0736
0.8388
0.0262
0.8189
0.2416
+Few-shot
0.5586
0.0284
1.8145
0.0115
0.6166
0.0505
0.7729
0.0762
0.7154
0.0245
0.8441
0.2558
+TTA
0.5532
0.0294
1.6273
0.0109
0.6521
0.0661
0.7219
0.0661
0.7609
0.0258
0.8182
0.2291
+SWIFT
0.5459
0.0278
1.5938
0.0094
0.5733
0.0432
0.7163
0.0686
0.7922
0.0260
0.8246
0.2415
TimesFM-2.5
0.9127
0.0414
1.9123
0.0119
0.5922
0.0532
0.8257
0.0814
0.8381
0.0284
0.7943
0.2320
Table 2: SWIFT across diverse TSFM backbones. Forecasting performance of zero-shot inference, few-shot fine-tuning, TTA, and SWIFT on the long-history subset of Chronos Benchmark II . The backbone-name rows report zero-shot results, while the remaining rows report the corresponding adaptation methods. Lower values indicate better performance. The best and second-best results within each backbone are marked in bold and underline , respectively.
Figure 3: Pairwise win rates across four TSFM backbones and six long-history datasets from Chronos Benchmark II . Each off-diagonal cell reports the percentage and count of comparisons in which the row method achieves a lower error than the column method. Panels (a) and (b) summarize 24 comparisons for MASE and WQL, respectively. Panel (c) pools the two metrics into 48 comparisons, rather than requiring simultaneous improvements in both metrics. Darker colors indicate higher win rates. Diagonal entries are omitted.
Benchmark Subset
# Tasks
Relative Improvement
WQL
MASE
Chronos Benchmark II
6
+6.99%
+6.31%
fev-bench
46
+1.66%
+2.14%
GIFT-Eval
24
+1.23%
+1.25%
Table 3: SWIFT across forecasting benchmarks. Average relative improvements over Chronos-2 zero-shot inference on history-qualified tasks satisfying T≥2048 .
Backbone
Method
Traffic
Electricity
Exchange
ETTm
ETTh
ERCOT
Avg.
Chronos-2
Few-shot
3.708
55.732
1.437
16.262
3.722
36.766
19.604
TTA
0.661
3.041
1.609
1.153
1.007
1.748
1.537
SWIFT
2.748
5.750
4.712
2.236
2.953
4.996
3.899
TimesFM-2.5
Few-shot
9.333
135.668
3.293
40.118
9.022
90.874
48.051
TTA
1.403
9.647
6.081
4.431
4.563
6.535
5.443
SWIFT
1.367
8.395
3.811
4.487
4.420
8.138
5.103
Table 4: Adaptation and forecasting latency across TSFM backbones. Model-side execution time in seconds for few-shot fine-tuning, TTA, and SWIFT on the long-history subset of Chronos Benchmark II . Few-shot latency includes dataset-level LoRA fine-tuning and subsequent forecasting. TTA latency includes ten sequential GCM updates on preceding pseudo-windows and one complete target-window forecast. SWIFT latency includes support-set construction, 15 LoRA updates, base and adapted forecasting, and prediction fusion. Model loading, dataset downloading, and data-availability waiting time are excluded. All results are rounded to three decimal places.
Setting
MASE Imp.
WQL Imp.
Zero-shot
0.00%
0.00%
Full SWIFT
+6.31%
+6.99%
w/o Anchor
+6.00%
+6.19%
w/o Frequency
+4.40%
+5.04%
w/o Fusion
+6.32%
+5.90%
w/o RAG
+2.05%
+2.38%
Table 5: Component ablation of SWIFT on Chronos Benchmark II . MASE and WQL improvements are measured relative to Chronos-2 zero-shot inference. All adapted variants use 20 adaptation pairs and 15 gradient steps, with all non-ablated settings held fixed. Dataset-wise relative improvements are computed first and then averaged across the six long-history datasets. The best result for each metric is marked in bold.
Time series forecasting predicts future values from ordered observations using statistical, neural, and probabilistic models to support decisions under uncertainty.
Inspired by recent breakthroughs in large language models for natural language processing, foundation models have emerged as a promising paradigm for zero-shot time series forecasting, enabling accurate predictions on datasets never seen during pre-training. Ranging from tens to hundreds of millions of parameters, these models are pre-trained on vast and diverse collections of time series, learning generalizable representations that support both point and probabilistic forecasting. This approach alleviates the need for dataset-specific model design and manual tuning, offering a unified solution across forecasting problems. In this work, we review the main architectures, pre-training strategies, and optimization methods underpinning these models. We further investigate post-pre-training fine-tuning of selected foundation models to enhance their performance on specific datasets. Our empirical results demonstrate that this step consistently improves forecasting accuracy over the zero-shot baseline.
Morad Laglil, Bertrand Pracca, Emilie Devijver +1
Savoye, 18 bd des Gorgets, 21000, Dijon, France · Univ. Grenoble Alpes, CNRS, Grenoble INP, LIG, Grenoble, 38000, France
Large Time Series Foundation Models (TSFMs) demonstrate strong zero-shot forecasting capabilities across diverse domains. However, their application to regional climate forecasting faces practical challenges: model weights are often proprietary, local training records are limited, and computational budgets are constrained, making traditional fine-tuning approaches infeasible. To address these constraints, we introduce a lightweight, black-box adaptation framework (requiring no access to backbone parameters and no backbone fine-tuning) that enhances frozen TSFMs at inference time through two plug-and-play wrappers: SMR\textsuperscript{2} (Stationarity aware multi-resolution Residual), which decomposes the input into multi-resolution temporal views, learns stride specific residual corrections that capture regional dynamics, then adaptively ensembles them into a single forecast, and MBB (Moving Block Bootstrap), which preserves temporal dependencies through block resampling and ensembles over temporally coherent residual perturbations to stabilize the point forecast. Both wrappers instantiate the same bagging style principle: they build diverse views of the input or its residuals, forecast each with the same frozen backbone, and aggregate, so all adaptation comes from inference time ensembling rather than any weight update. Evaluated on one month ahead Standardized Precipitation Evapotranspiration Index (SPEI) prediction across multiple sites in South Australia, our framework consistently improves forecasting performance across several backbone models, demonstrating up to 26% mean squared error (MSE) reduction over the corresponding frozen backbone while enabling practical deployment in resource constrained regional forecasting systems.
Wentao Gao, Jiuyong Li, Lin Liu +4
Adelaide University · Adelaide, Australia · CSIRO Technology +1
Time Series Foundation Models (TSFMs) advance generalization and data efficiency in time series forecasting by unified large-scale pretraining. But TSFMs remain lacking when adapting to specific downstream forecasting tasks for two reasons. First, the non-stationary and uncertain nature of time series data lead to inevitable temporal distribution shifts between historical training and future testing data, while current Supervised FineTuning (SFT)-based methods are prone to overfitting and may degrade generalization. Second, training data availability varies across forecasting tasks, requiring TSFMs to generalize well under diverse data regimes. To address these challenges, we introduce the Time series Reinforcement Finetuning (TimeRFT) paradigm for TSFM downstream adaptation, which consists of two task-specific training recipes: i) A forecasting quality-based temporal reward mechanism that conducts a multi-faceted evaluation of the contribution of each prediction step to overall forecasting accuracy. ii) A forecasting difficulty-based data selection strategy to identify time series samples with generalizable predictive patterns and informative training signals. Extensive experiments demonstrate TimeRFT can consistently outperform SFT-based adaptation methods across various real-world forecasting tasks and training data regimes, enhancing prediction accuracy and generalization against unforeseen distribution shifts.
Siyang Li, Yize Chen, Zijie Zhu +4
HKUST(GZ) Guangzhou, China · University of Alberta Edmonton, Canada · Alibaba Cloud Guangzhou, China +3