Kun Gai

MemLearner achieves unprecedented scene consistency in video generation by learning to query context memory, outperforming traditional methods in dynamic and occluded environments.

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

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

NormGuard effectively preserves reward alignment in RL fine-tuning while significantly enhancing perceptual quality, challenging the notion that post-training improvements come at the cost of visual fidelity.

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

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

Generating personalized videos on demand can boost ad revenue by nearly 2% in high-traffic environments, transforming traditional recommendation systems.

A novel approach that boosts LTV prediction for billions of low-activity users by transforming sparse profiles into actionable insights without heavy LLM reliance.

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