Framework separates perception and reasoning for time series question answering
Zero-Shot Time-Series Question Answering via Decoupled Perception and Reasoning
Artificial Intelligence
Summary
Answering questions about time-based data is tough because the data and questions can vary a lot. The authors propose TSHarness, a system that first extracts important numbers and patterns from the data, then reasons about those to answer questions. By separating these steps, the system can handle new kinds of questions and data without extra training. This makes it more flexible and efficient for understanding time series information.
What this means in practice
- •For business analysts: Interpret complex financial or sales time-series data through natural language queries without retraining models.
- •For energy monitoring teams: Quickly analyze sensor data streams by asking flexible questions about temporal trends and statistics on new datasets.
Authors
Jing Xie, Haochen Yuan, Yunbo Wang
Abstract
Time-series question answering (TSQA) requires grounding linguistic queries and diverse answer formats in complex numerical observations. However, existing methods heavily overfit to specific datasets and struggle to generalize when input series, question contexts, and answer requirements shift simultaneously. To address this challenge, we propose TSHarness, an agentic framework that establishes a decoupled workflow for cross-dataset zero-shot TSQA. At its core, TSHarness divides and conquers numerical perception and contextual reasoning via a structured Time-Series Perception State (TPS). Guided by a reusable memory of analytical knowledge, a learned Tool Selector adaptively invokes numerical tools to extract salient statistical and temporal features into the TPS. The answering agent then performs semantic reasoning over the TPS to generate target outputs, triggering iterative re-perception through the feedback loop when evidence is deemed insufficient. By separating numerical feature extraction from question-specific reasoning, TSHarness eliminates the need for target-side training or answer feedback, providing a generalizable, cost-efficient foundation for zero-shot TSQA.