LEMs: A Primer On Large Execution Models

2025-09-21

AI Summary

This paper addresses the complex trade execution problem characterized by flexible time horizons and multiple execution constraints. We propose Large Execution Models (LEMs), a unified framework extending the Transformer architecture. Methodologically, LEMs introduce a novel formulation of execution duration as an adjustable min-max time window; decouple market-state representation from execution-policy logic to enable cross-scenario feature sharing and independent allocation; and integrate Temporal Kolmogorov–Arnold Networks (KANs), variable selection networks, multi-head attention, and a dedicated allocation network for dynamic path optimization. Empirically, LEMs achieve statistically significant outperformance over traditional benchmarks—namely, VWAP, TWAP, and reinforcement learning baselines—in both intraday cryptocurrency trading and multi-day execution across Dow Jones Industrial Average constituents. Results demonstrate LEMs’ robustness and generalizability under dynamic temporal constraints, establishing a new state-of-the-art for adaptive algorithmic execution.

Technology Category

• Search and Optimization: Algorithm Configuration
• Machine Learning: Mixture of Experts (MoE)
• Multiagent Systems: Mechanism Design

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

This paper introduces Large Execution Models (LEMs), a novel deep learning framework that extends transformer-based architectures to address complex execution problems with flexible time boundaries and multiple execution constraints. Building upon recent advances in neural VWAP execution strategies, LEMs generalize the approach from fixed-duration orders to scenarios where execution duration is bounded between minimum and maximum time horizons, similar to share buyback contract structures. The proposed architecture decouples market information processing from execution allocation decisions: a common feature extraction pipeline using Temporal Kolmogorov-Arnold Networks (TKANs), Variable Selection Networks (VSNs), and multi-head attention mechanisms processes market data to create informational context, while independent allocation networks handle the specific execution logic for different scenarios (fixed quantity vs. fixed notional, buy vs. sell orders). This architectural separation enables a unified model to handle diverse execution objectives while leveraging shared market understanding across scenarios. Through comprehensive empirical evaluation on intraday cryptocurrency markets and multi-day equity trading using DOW Jones constituents, we demonstrate that LEMs achieve superior execution performance compared to traditional benchmarks by dynamically optimizing execution paths within flexible time constraints. The unified model architecture enables deployment across different execution scenarios (buy/sell orders, varying duration boundaries, volume/notional targets) through a single framework, providing significant operational advantages over asset-specific approaches.

Problem

• Extending transformer architectures for complex execution problems with flexible time boundaries
• Decoupling market information processing from execution allocation decisions across scenarios
• Optimizing execution paths dynamically within flexible time constraints for superior performance

Innovation

• Transformer-based deep learning for flexible execution problems
• Decouples market data processing from allocation decisions
• Unified model handles diverse execution scenarios through single framework