TradeFM: A Generative Foundation Model for Trade-flow and Market Microstructure

Researchers introduced TradeFM, a 524M-parameter generative AI model that learns from billions of trade events across 9,000+ equities to understand market microstructure. The model can generate synthetic market data and generalizes across different markets without asset-specific calibration, potentially enabling new applications in trading and market simulation.

• →TradeFM is a foundation model trained on billions of trade events from over 9,000 equities using a 524M-parameter Transformer architecture.
• →The model uses scale-invariant features and universal tokenization to work across different assets without specific calibration.
• →TradeFM achieves 2-3x lower distributional error than existing Compound Hawkes baselines in market simulation.
• →The model successfully generalizes to out-of-distribution APAC markets with only moderate performance degradation.
• →Applications include synthetic data generation, stress testing, and development of learning-based trading agents.