Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

TimeART: Towards Agentic Time Series Reasoning via Tool-Augmentation

Xingjian Wu, Junkai Lu, Zhengyu Li, Xiangfei Qiu, Jilin Hu, Chenjuan Guo, Christian S. Jensen, Bin Yang

TL;DR

TimeART tackles automation in time series analysis by fusing LLM-based reasoning with tool augmentation. It introduces TimeToolBench, a 100k trajectory corpus, and a four-stage training strategy to teach TSRMs strategic, self-reflective tool-use. Experiments with an 8B TSRM trained on TimeToolBench show state-of-the-art performance on TSQA benchmarks MTBench and TimeMQA, approaching or matching larger models with far fewer parameters. The work demonstrates that combining external analytical tools with agentic reasoning substantially improves accuracy and generalization in time-series question answering, with practical impact on cyber-physical systems analytics.

Abstract

Time series data widely exist in real-world cyber-physical systems. Though analyzing and interpreting them contributes to significant values, e.g, disaster prediction and financial risk control, current workflows mainly rely on human data scientists, which requires significant labor costs and lacks automation. To tackle this, we introduce TimeART, a framework fusing the analytical capability of strong out-of-the-box tools and the reasoning capability of Large Language Models (LLMs), which serves as a fully agentic data scientist for Time Series Question Answering (TSQA). To teach the LLM-based Time Series Reasoning Models (TSRMs) strategic tool-use, we also collect a 100k expert trajectory corpus called TimeToolBench. To enhance TSRMs' generalization capability, we then devise a four-stage training strategy, which boosts TSRMs through learning from their own early experiences and self-reflections. Experimentally, we train an 8B TSRM on TimeToolBench and equip it with the TimeART framework, and it achieves consistent state-of-the-art performance on multiple TSQA tasks, which pioneers a novel approach towards agentic time series reasoning.

TimeART: Towards Agentic Time Series Reasoning via Tool-Augmentation

TL;DR

TimeART tackles automation in time series analysis by fusing LLM-based reasoning with tool augmentation. It introduces TimeToolBench, a 100k trajectory corpus, and a four-stage training strategy to teach TSRMs strategic, self-reflective tool-use. Experiments with an 8B TSRM trained on TimeToolBench show state-of-the-art performance on TSQA benchmarks MTBench and TimeMQA, approaching or matching larger models with far fewer parameters. The work demonstrates that combining external analytical tools with agentic reasoning substantially improves accuracy and generalization in time-series question answering, with practical impact on cyber-physical systems analytics.

Abstract

Time series data widely exist in real-world cyber-physical systems. Though analyzing and interpreting them contributes to significant values, e.g, disaster prediction and financial risk control, current workflows mainly rely on human data scientists, which requires significant labor costs and lacks automation. To tackle this, we introduce TimeART, a framework fusing the analytical capability of strong out-of-the-box tools and the reasoning capability of Large Language Models (LLMs), which serves as a fully agentic data scientist for Time Series Question Answering (TSQA). To teach the LLM-based Time Series Reasoning Models (TSRMs) strategic tool-use, we also collect a 100k expert trajectory corpus called TimeToolBench. To enhance TSRMs' generalization capability, we then devise a four-stage training strategy, which boosts TSRMs through learning from their own early experiences and self-reflections. Experimentally, we train an 8B TSRM on TimeToolBench and equip it with the TimeART framework, and it achieves consistent state-of-the-art performance on multiple TSQA tasks, which pioneers a novel approach towards agentic time series reasoning.
Paper Structure (19 sections, 7 equations, 7 figures, 4 tables)

This paper contains 19 sections, 7 equations, 7 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Methodology
  3. TimeART
  4. TimeToolBench
  5. Training strategy
  6. Experiments
  7. Experimental Settings
  8. Main Results
  9. Model Analysis
  10. Related Works
  11. LLMs for Time Series Analysis
  12. Agents for Time Series Analysis
  13. Conclusion
  14. Implementation details
  15. Dataset Details
  16. ...and 4 more sections

Figures (7)

  • Figure 1: Common dilemmas in time series reasoning, i.e., numerical hallucination and cognitive defect.
  • Figure 2: Common dilemmas in behaviour cloning and reinforcement learning, causing low generalization and entropy collapse.
  • Figure 3: Ratios of data sources in TimeToolBench (100k), the ReAct-style training corpus for agentic time series reasoning.
  • Figure 4: The overviews of the data pipeline for TimeToolBench (left), and the training pipeline for TimeART (right).
  • Figure 5: Ablations on the TimeART framework.
  • ...and 2 more figures