MemPromptTSS: Persistent Prompt Memory for Iterative Multi-Granularity Time Series State Segmentation

2025-10-10

AI Summary

Existing time-series segmentation prompting methods rely solely on local context, leading to rapid prompt-effect decay and limiting cross-sequence multi-granularity state identification and iterative optimization. To address this, we propose a persistent prompt memory mechanism that encodes prompts together with their neighboring subsequences into reusable memory tokens, forming a cross-sequence prompt memory bank. Integrated with multi-granularity attention, our method jointly models local cues and historical prompts at each segmentation step, enabling unified representation of coarse-grained patterns and fine-grained events. Evaluated on six wearable and industrial monitoring datasets, our approach achieves an 85% improvement in multi-granularity segmentation accuracy over the best baseline under single-pass inference, and yields an average per-iteration gain of 2.66 percentage points during iterative inference—substantially outperforming PromptTSS (1.19).

Technology Category

• Natural Language Processing: Prompt Engineering / Prompting
• Computer Vision: Segmentation
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Search and Retrieval-Augmented AI: Personalized, context-aware and across-device search
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Web platforms, mobile applications, and connected sensing systems generate multivariate time series with states at multiple levels of granularity, from coarse regimes to fine-grained events. Effective segmentation in these settings requires integrating across granularities while supporting iterative refinement through sparse prompt signals, which provide a compact mechanism for injecting domain knowledge. Yet existing prompting approaches for time series segmentation operate only within local contexts, so the effect of a prompt quickly fades and cannot guide predictions across the entire sequence. To overcome this limitation, we propose MemPromptTSS, a framework for iterative multi-granularity segmentation that introduces persistent prompt memory. A memory encoder transforms prompts and their surrounding subsequences into memory tokens stored in a bank. This persistent memory enables each new prediction to condition not only on local cues but also on all prompts accumulated across iterations, ensuring their influence persists across the entire sequence. Experiments on six datasets covering wearable sensing and industrial monitoring show that MemPromptTSS achieves 23% and 85% accuracy improvements over the best baseline in single- and multi-granularity segmentation under single iteration inference, and provides stronger refinement in iterative inference with average per-iteration gains of 2.66 percentage points compared to 1.19 for PromptTSS. These results highlight the importance of persistent memory for prompt-guided segmentation, establishing MemPromptTSS as a practical and effective framework for real-world applications.

Problem

• Persistent prompt memory enables iterative multi-granularity time series segmentation
• Overcomes fading prompt effects by maintaining influence across entire sequences
• Integrates domain knowledge through persistent memory tokens for improved accuracy

Innovation

• Persistent prompt memory enables iterative multi-granularity segmentation
• Memory encoder stores prompt tokens for cross-sequence conditioning
• Framework maintains prompt influence across entire time series