Publications

(2026). GPT4D: Generative Pre-training Transformer with Next-Scale Spatio-temporal Token Prediction for 4D Human Action Recognition.