Problem
Research questions and friction points this paper is trying to address.
Time-Series Foundation Models
Inference Cost
Context Efficiency
Forecasting
Instance-Adaptive Prompts as Context for Time-Series Foundation Models
🤖 AI Summary
This study addresses the prohibitive inference costs incurred by long historical inputs in time-series foundation models. To this end, we propose PaCTS, a method that generates instance-adaptive latent prompts as compact context substitutes for frozen models. By integrating global statistics with local temporal information, PaCTS constructs continuous embedding tokens and introduces a visible-context-conditioned learning strategy coupled with a joint training mechanism to achieve efficient context representation. Experimental results demonstrate that PaCTS, utilizing substantially shorter inputs, surpasses the predictive performance of baselines employing twice the context length. Furthermore, it significantly reduces computational overhead while effectively enhancing out-of-distribution generalization capabilities.
📝 Abstract
Longer histories can improve time-series foundation models (TSFMs), but require substantially higher inference cost. We therefore ask whether contextual information can be provided more efficiently through a compact set of learned token embeddings. We introduce PaCTS, which generates a small set of instance-adaptive latent prompts in the form of continuous embedding tokens conditioned on the visible context. These prompts serve as compact context surrogates for frozen TSFMs. PaCTS constructs them from instance-specific global statistics and further refines them with segment-level temporal information, capturing both global characteristics and local temporal variations. The prompt module is jointly trained and deployed across heterogeneous time series with the frozen backbone. Extensive experiments demonstrate the effectiveness of prompts as context, consistently improving forecasting across context lengths and model architectures. With a shorter input context, PaCTS can outperform the same frozen backbone using double context while requiring substantially less inference computation. Compared with weight-space adaptation methods, PaCTS achieves stronger improvements and better out-of-distribution generalization.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Instance-Adaptive Prompts
Time-Series Foundation Models
Continuous Embedding Tokens
Context Surrogates
Out-of-Distribution Generalization
🔎 Similar Papers
No similar papers found.
