arXiv 2602.23784
TradeFM: A Generative Foundation Model for Trade-flow and Market Microstructure
By Maxime Kawawa-Beaudan, Srijan Sood, et al.
Published 2026-02-27
Citation lineage
Review the prior work and downstream research connected to this paper.
Foundation models have transformed domains from language to genomics by learning general-purpose representations from large-scale, heterogeneous data. We introduce TradeFM, a 524M-parameter generative Transformer that brings this paradigm to market microstructure, learning directly from billions of trade events across >9K equities. To enable cross-asset generalization, we develop scale-invariant features and a unive…