Yi Li

OneRanker tackles the optimization tension in generative recommendation systems by deeply integrating generation and ranking into a single architecture, achieving state-of-the-art performance in industrial advertising.

Personalized federated learning gets a boost from Fed-GAME, which uses a graph attention mixture-of-experts to dynamically aggregate client updates, outperforming SOTA methods on time-series forecasting tasks.

Key contribution not extracted.

Standard multimodal LLMs can perform surprisingly well on dense prediction tasks like segmentation and depth estimation, without needing any task-specific decoder modules.