Ruiming Tang

Generative retrieval can achieve both shared modeling and objective-specific control, leading to significant improvements in user engagement metrics.

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

IID-Nav achieves high-precision deep retrieval by enabling logical unlimited-depth graph traversal without increasing inference latency.

Real-time recommendation can be transformed by capturing dynamic user interests through a novel partial-order modeling approach that significantly boosts engagement metrics.

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

UniFormer achieves up to a 1.113% increase in Watch Time, showcasing a breakthrough in user engagement for industrial recommendation systems.

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.

Domain-specialized LLMs can regain lost general skills without sacrificing their expertise, thanks to a new distillation method that disentangles conflicting training signals.

A2Gen transforms short video recommendations by treating user actions as dynamic sequences, resulting in substantial improvements in user engagement metrics.

Sub-linear attention is now possible without sacrificing complete long-range dependency retention, thanks to learnable summary tokens that compress context.

Generative recommendation models like OneRec-V2 can achieve near-lossless FP8 quantization, unlocking significant latency and throughput improvements, unlike traditional recommender systems.