autoregressive diffusion models

The trading firm Jane Street has published a blog post asking whether autoregressive diffusion models can be used to generate synthetic market data. The piece, written by one of its engineers, examines how generative techniques developed for images and text might produce realistic financial time series. Readers are debating the technical merits and possible trading applications.

Trading firm Jane Street has published a blog post examining whether autoregressive diffusion models can be used to generate synthetic market data. The piece walks through how these generative models, best known for image and text generation, might produce realistic financial time series. It is drawing attention among machine learning practitioners and quants discussing generative approaches to market simulation.