# Pseudo-Label-Triggered Retraining from Forecast Errors for Online Time Series Forecasting

arXiv: https://arxiv.org/abs/2609.39789 · HTML: https://mlsanity.com/paper/2609.39789/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 8 of its 32 references that are in the map.
- First posted on arXiv 30 Sep 2026.

## Builds on (most cited first, 8 of 8)

- [Informer: Beyond Efficient Transformer for Long Sequence Time-Series Forecasting](https://mlsanity.com/paper/2012.07436/) (2020)
- [Autoformer: Decomposition Transformers with Auto-Correlation for Long-Term Series Forecasting](https://mlsanity.com/paper/2106.13008/) (2021)
- [Learning under Concept Drift: A Review](https://mlsanity.com/paper/2004.05785/) (2020)
- [Online Learning: A Comprehensive Survey](https://mlsanity.com/paper/1802.02871/) (2018)
- [Addressing Concept Shift in Online Time Series Forecasting: Detect-then-Adapt](https://mlsanity.com/paper/2403.14949/) (2024)
- [Handling Concept Drift in Global Time Series Forecasting](https://mlsanity.com/paper/2304.01512/) (2023)
- [On the retraining frequency of global models in retail demand forecasting](https://mlsanity.com/paper/2505.00356/) (2025)
- [Online detection of forecast model inadequacies using forecast errors](https://mlsanity.com/paper/2502.14173/) (2025)

## Abstract (arXiv)

Real-world time series forecasting systems operate under non-stationary data streams, where forecasting performance may degrade over time. Although retraining can recover the performance, it incurs non-trivial computational and operational costs. Under limited deployment resources, the key challenge is therefore not only how to retrain but also when to retrain. While existing retraining policies often rely on indirect indicators such as drift alarms or model staleness, we instead use realized forecast errors as direct deployment feedback. In this paper, we propose PILOT (Pseudo-label-Informed Learned Online Trigger), an online retraining framework that learns when to retrain from forecast-error dynamics. Since ground-truth retraining labels are unavailable, PILOT constructs a pseudo-label from future increases in forecast error and trains a lightweight scorer to predict it from observed error states. At deployment, PILOT uses only completed forecast errors and serves as a plug-in module for arbitrary forecasting backbones without architectural modification. We evaluate PILOT under standard multivariate forecasting settings across eight benchmarks with three representative backbones---DLinear, iTransformer, and TimesNet. Across all three backbones, PILOT achieves state-of-the-art average-rank performance among retraining policies while maintaining a favorable performance--efficiency trade-off.