AdaCast: Conditional Parameter Generation for Adaptive Time Series Forecasting

2026-10-08

AI Summary

This study addresses the limitation of existing time-series foundation models that rely on static adaptation, which prevents parameter adjustment based on input dynamics and constrains both personalized forecasting and generalization. To overcome this, we propose AdaCast, a conditional parameter generation framework that keeps the pre-trained model frozen while employing a generator to dynamically synthesize low-rank parameter updates tailored to each input, enabling adaptive inference. This work introduces the first input-dependent dynamic conditional parameter generation mechanism, surpassing the constraints of conventional static full fine-tuning. Extensive evaluations across six public benchmarks demonstrate that AdaCast consistently outperforms static baselines, yielding substantial improvements in both in-domain forecasting accuracy and cross-domain zero-shot generalization.

Abstract

Time-series foundation models (TSFMs) have achieved strong forecasting performance across domains. However, most adaptation methods remain static. Existing all-in-one methods learn a single set of dataset-level parameter updates and apply the same adapted model to every input. As a result, they cannot adapt the model parameters to the temporal patterns, seasonality and dynamics of each input time series. This limits their ability to produce forecasts that are tailored to heterogeneous inputs. To address this limitation, we propose AdaCast, a conditional parameter generation framework for time-series forecasting. AdaCast uses a generator to produce input-specific low-rank parameter updates for a frozen pretrained TSFM. These updates adapt the model to each input during both training and inference. Across six public benchmarks, AdaCast consistently outperforms static adaptation baseline in in-domain forecasting and improves zero-shot generalization to held-out datasets across domains. These results demonstrate that conditional parameter generation provides an effective approach for adaptive forecasting.

Problem

• Time-series foundation models
• Adaptive forecasting
• Static adaptation
• Heterogeneous inputs
• Conditional parameter generation

Innovation

• Conditional Parameter Generation
• Time-Series Foundation Models
• Low-Rank Adaptation
• Adaptive Forecasting
• Zero-Shot Generalization