FinSTaR: Towards Financial Reasoning with Time Series Reasoning Models

2026-05-05

AI Summary

This work addresses the challenge that existing time series reasoning models struggle to effectively handle the dual demands of deterministic evaluation and stochastic forecasting inherent in financial tasks. To bridge this gap, the authors introduce FinTSR-Bench, the first benchmark specifically designed for financial time series reasoning, along with FinSTaR, a novel model that unifies programmatic chain-of-thought (Compute-in-CoT) and scenario-aware chain-of-thought (Scenario-Aware CoT) mechanisms. Built upon a proposed 2×2 taxonomy of financial reasoning capabilities, FinSTaR achieves an average accuracy of 78.9% across ten diverse financial reasoning tasks on FinTSR-Bench, substantially outperforming both state-of-the-art large language models and conventional time series baselines. These results demonstrate the complementary benefits and efficacy of jointly training multiple reasoning capabilities within a unified framework.

Abstract

Time series (TS) reasoning models (TSRMs) have shown promising capabilities in general domains, yet they consistently fail on financial domain, which exhibit unique characteristics. We propose a general 2x2 capability taxonomy for TSRMs by crossing 1) single-entity vs. multi-entity analysis with 2) assessment of the current state vs. prediction of future behavior. We instantiate this taxonomy in the financial domain -- where the distinction between deterministic assessment and stochastic prediction is particularly critical -- as ten financial reasoning tasks, forming the FinTSR-Bench benchmark based on S&P stocks. To this end, we propose FinSTaR (Financial Time Series Thinking and Reasoning), trained on FinTSR-Bench with distinct chain-of-thought (CoT) strategies tailored to each category. For assessment, which is deterministic (i.e., computable from observable data), we employ Compute-in-CoT, a programmatic CoT that enables models to derive answers directly from raw prices. For prediction, which is inherently stochastic (i.e., subject to unobservable factors), we adopt Scenario-Aware CoT, which generates diverse scenarios before making a judgment, mirroring how financial analysts reason under uncertainty. The proposed method achieves 78.9% average accuracy on FinTSR-Bench, substantially outperforming LLM and TSRM baselines. Furthermore, we show that the four capability categories are complementary and mutually reinforcing through joint training, and that Scenario-Aware CoT consistently improves prediction accuracy over standard CoT. Code is publicly available at: https://github.com/seunghan96/FinSTaR.

Problem

• financial reasoning
• time series reasoning
• deterministic assessment
• stochastic prediction
• multi-entity analysis

Innovation

• Time Series Reasoning
• Chain-of-Thought
• Financial Reasoning
• Scenario-Aware Reasoning
• Compute-in-CoT