PaP-NF: Probabilistic Long-Term Time Series Forecasting via Prefix-as-Prompt Reprogramming and Normalizing Flows
The article discusses a new probabilistic forecasting framework called PaP-NF, which is designed for long-term time series forecasting. This framework utilizes a Prefix-as-Prompt mechanism to align time series data with a large language model and employs a normalizing flow decoder to capture uncertainty. The authors demonstrate that PaP-NF effectively maintains competitive accuracy while robustly representing multi-modal uncertainty in predictions.
- ▪PaP-NF is a probabilistic forecasting framework for long-term time series data.
- ▪It uses a Prefix-as-Prompt mechanism to align data with a frozen large language model.
- ▪The framework captures multi-modal uncertainty while maintaining competitive point forecasting accuracy.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Machine Learning arXiv:2605.23219 (cs) [Submitted on 22 May 2026] Title:PaP-NF: Probabilistic Long-Term Time Series Forecasting via Prefix-as-Prompt Reprogramming and Normalizing Flows Authors:Minju Kim, Youngbum Hur View a PDF of the paper titled PaP-NF: Probabilistic Long-Term Time Series Forecasting via Prefix-as-Prompt Reprogramming and Normalizing Flows, by Minju Kim and 1 other authors View PDF HTML (experimental) Abstract:Time series forecasting plays a central role in many real-world applications and has been extensively studied. Most existing approaches rely on deterministic models. However, real-world environments exhibit inherently uncertain and complex future behaviors, making single-point predictions insufficient.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv cs.AI.
Discussion
0 commentsMore from arXiv cs.AI
