GoKawiilJane Street intern builds diffusion model to generate synthetic market data
A 2026 summer research intern at Jane Street, Kavish, developed a diffusion model aimed at synthesizing order book events—such as order placements, cancellations and executions—rather than just predicting prices. The project explored whether market data, which mixes discrete actions with near-continuous and often spiky distributions (like order timing and price ticks), can be modeled using diffusion techniques borrowed from image, video, and audio generation.
GoKawiil's interpretation of the reporting above, not reported fact.
Treating market microstructure as a generative modeling problem could eventually allow richer simulations of trading activity, giving researchers full synthetic order flow rather than single price forecasts. The firm's framing suggests this is an exploratory research effort to understand where continuous-data assumptions break down for inherently discrete, bursty market events, which could inform future modeling approaches in quantitative finance.
- A Jane Street summer intern built a diffusion model to generate synthetic order book data.
- The project treats market data as continuous, testing where that assumption fails given discrete actions and spiky timing/price distributions.
- The approach borrows techniques from diffusion and flow-matching models used in image, video, audio, and robotics generation.
Source: blog.janestreet.com, 2026-10-09
Published there as: “Can you use autoregressive diffusion to generate market data?”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.