Shuang Yang

RGD reshapes the decoding process in generative recommendation, ensuring high-value candidates are prioritized without retraining the model.

Behavior-derived rewards enable MLLMs to generate effective recommendations without user-specific data at inference, outperforming traditional methods.

A unified model that seamlessly integrates generative recall and multi-objective ranking achieves significant performance improvements in real-world recommendation systems.

Diffusion models can now generate user preferences for multi-behavior sequential recommendation, outperforming traditional methods by better capturing uncertainty and enabling more diverse recommendations.

Achieve lossless acceleration of ranking models by structurally re-parameterizing feature fusion matrix multiplication, sidestepping the accuracy drop common in lightweighting and distillation.