Fei Pan

LGRID transforms the landscape of local-life service recommendations by generating interpretable Semantic IDs that are both semantically rich and collision-resistant.

Bridging the Understanding-Action Gap leads to a 4.5% revenue increase by refining recommendation policies through direct user feedback rather than linguistic reasoning alone.

Latent reasoning can boost recommendation efficiency by over 10x while enhancing accuracy, challenging the need for verbose rationales in LLM applications.

Not every missing modality needs to be repaired for optimal sentiment analysis, and SIEVE learns to make this decision dynamically at the sample level.

AgentX can autonomously iterate on recommendation algorithms, outpacing human-driven processes and fundamentally changing how we approach system development.

Surprisingly, the "think before answer" paradigm fails to enhance generative recommendation models, prompting a novel approach that redefines how reasoning is integrated into these systems.

LLMs can master auto-bidding in dynamic ad environments, but only if you give them a hierarchical architecture and offline RL fine-tuning to avoid hallucinating suboptimal decisions.