Researchers release paper on LOBGrad, a conditional diffusion limit order book prediction framework

Junyu Chen, Franklin Liu, Moshu Xu and Lijian Yang published an academic paper proposing LOBGrad, a conditional DDPM framework for probabilistic limit order book prediction, which combines an autoregressive RNN encoder with a diffusion-based generator. Under exponential mixing and regularity assumptions, the paper establishes an excess-risk oracle inequality for the single-step DDPM noise prediction objective; under contraction dynamics, the dependence on the conditioning window length can be simplified to dependence on the latent state dimension. Experiments on BTC/USDT, GOOGL and META show that LOBGrad outperforms diffusion ablations without an encoder across all reported queue metrics, and yields competitive results compared with other baseline methods.

AI-generated from reports · 3 hr agoUpdated

Not enough consecutive observations yet; trend is not drawn.