Time Series Foundation Models Improve LLM Decisions: A Case Study in Stock Trading
Shifeng Xie ⋅ Ziwei Li ⋅ Themis Palpanas ⋅ Peilin Zhao ⋅ Chenghao Liu
Abstract
Time series foundation models provide strong numerical forecasting capabilities, while large language models are strong at high-level reasoning, suggesting a complementary synergy for decision-making tasks. We study this integration in a stock trading setting showing that TSFM signals statistically significant improvements in trading performance over LLM-only baselines.
Chat is not available.
Successful Page Load