PATRA: Pattern-Aware Alignment and Balanced Reasoning for Time Series Question Answering

TL;DR

PATRA is a novel model that enhances time series question answering by extracting patterns and balancing learning across tasks, leading to improved reasoning and understanding.

Contribution

The paper introduces PATRA, a pattern-aware and task-balanced approach that significantly improves deep reasoning in time series question answering.

Findings

PATRA outperforms existing baselines on diverse TSQA tasks.

It effectively extracts trend and seasonality patterns for better alignment.

The model demonstrates superior reasoning and cross-modal understanding.

Abstract

Time series reasoning demands both the perception of complex dynamics and logical depth. However, existing LLM-based approaches exhibit two limitations: they often treat time series merely as text or images, failing to capture the patterns like trends and seasonalities needed to answer specific questions; and when trained on a mix of simple and complex tasks, simpler objectives often dominate the learning process, hindering the development of deep reasoning capabilities. To address these limitations, we propose the Pattern-Aware Alignment and Balanced Reasoning model (PATRA), introducing a pattern-aware mechanism that extracts trend and seasonality patterns from time series to achieve deep alignment. Furthermore, we design a task-aware balanced reward to harmonize learning across tasks of varying difficulty, incentivizing the generation of coherent Chains of Thought. Extensive experiments show that PATRA outperforms strong baselines across diverse Time Series Question Answering (TSQA) tasks, demonstrating superior cross-modal understanding and reasoning capability.