In a new research paper, scientists explore the use of large language models (LLMs) for parent-order execution, a core problem in algorithmic trading. The goal is to split a large order into smaller ones while minimizing execution costs. Traditional methods either rely on market assumptions that may not hold in practice or require task-specific training, limiting their adaptability.
The proposed framework, called PACE (Plan-Ahead Controlled Execution), decomposes the problem into long-horizon planning and short-horizon execution. This hierarchical approach requires neither explicit market assumptions nor task-specific training, making it more versatile than existing solutions.
Experiments on Shenzhen Stock Exchange Level-1 data show that PACE outperforms TWAP, Almgren-Chriss, and learning-based baselines, exceeding the strongest baseline by 0.65 basis points. This marks the first systematic study of LLMs for parent-order execution, extending their use in finance from what to trade to how to execute.
Behavioral analysis reveals that LLMs make execution decisions differently from human investors: higher model confidence predicts better performance rather than worse returns, and the model trades earlier rather than procrastinating toward the deadline. These findings suggest that LLMs can complement human traders in execution decisions.