PACE Framework for LLM-Based Algorithmic Parent-Order Execution

The PACE framework applies LLMs to algorithmic trading by decomposing parent-order execution into hierarchical planning and execution stages. Tested on Shenzhen Stock Exchange data, it outperforms TWAP, Almgren-Chriss, and existing learning-based baselines without task-specific training.