Xiangjun Fan

The paper proposes DAG-MoE, a novel sparse Mixture-of-Experts framework that replaces standard weighted-sum aggregation with structural aggregation to enhance model performance and enable multi-step reasoning.

OmniOPD introduces a logit-free, chunk-level distillation framework that improves on standard On-Policy Distillation by using semantic similarity and peak-entropy scheduling, achieving state-of-the-art performance even with black-box teachers.

This paper proposes G2Rec, a scalable framework for industrial-scale generative recommendation that unifies graph-based user co-engagement modeling and semantic tokenization.

The paper introduces a memory agent to improve decision-making in long-horizon tasks by actively updating and intervening with reminders.