Innovation
Methods, ideas, or system contributions that make the work stand out.
Test-Time Adaptation
Time-Series Foundation Models
Residual-Aware Correction
Neighbor Retrieval
Cross-Backbone Transferability
RACE: Residual-Aware Test-Time Adaptation for Neighbor-Rich Time-Series Foundation Model Forecasting
🤖 AI Summary
This study addresses the prediction limitations of time series foundation models in neighbor-rich scenarios, where residual conflicts and cross-model pattern discrepancies degrade performance. We propose a two-stage test-time adaptation framework that first achieves training-free correction aggregation by aligning neighborhood residuals via retrieval, and subsequently introduces a lightweight gating mechanism to dynamically determine when corrections should be applied. This approach enhances predictive accuracy without modifying the backbone network while maintaining cross-architecture transferability. Experimental results demonstrate that the proposed framework significantly improves three evaluation metrics across four foundation models, yielding particularly pronounced gains on high-error queries. Furthermore, comprehensive comparisons across 72 cross-backbone configurations consistently outperform frozen target models.
📝 Abstract
Time-series foundation models (TSFMs) perform strongly across forecasting tasks, but their per-series inference is ill-suited to neighbor-rich forecasting, where each query has access to related but nonidentical historical series. Continuous glucose monitoring (CGM) and Web/cloud workloads exemplify this setting: CGM trajectories share physiological patterns but vary across individuals, devices, and conditions, while Web/cloud workloads combine common operating regimes with non-stationarity, heavy tails, and bursts. These histories share useful structure, yet neighbors are not equally relevant. Existing methods either fine-tune TSFMs for each target domain, incurring additional costs and offering limited transferability across backbones, or append retrieved series without verifying whether they support the current forecast. The key challenges are conflicting residual evidence from neighboring series and residual patterns that vary across TSFMs and forecasting tasks. We formulate test-time neighborhood scaling: using same-domain neighbor evidence without modifying the backbone. We propose RACE (Residual-Aware Correction of Forecasting Errors), a two-stage framework for using historical neighbors. We first retrieve query-compatible neighbors, align their residuals to the query scale, and aggregate coherent evidence into the training-free RACE-TF correction. Full RACE then uses a lightweight, domain-specific Gate to determine when applying the correction is beneficial, with a reusable training workflow across TSFM backbones. Across four TSFMs, RACE improves all three domain-aggregate metrics on both primary domains, with the largest gains on high-error queries. Within each domain, a Gate trained on one TSFM transfers to other backbones without adaptation, and the resulting pipeline improves all 72 cross-backbone metric comparisons over the matched frozen targets.
🔎 Similar Papers
No similar papers found.
