Yonghui Yang

Hallucination mitigation in LVLMs doesn't have to come at the cost of general performance: MPD reduces hallucinations by 23.4% while *improving* overall generative capabilities.

Stop reinforcing redundant information: CLEAR boosts multimodal recommendation by projecting representations onto a null space that suppresses shared semantics across modalities.

Denoising diffusion models can disentangle user preferences from noisy micro-video interactions, leading to better recommendations.

Mamba's global context modeling gives YOLO a boost in murky underwater object detection, improving mAP by 4.9% over YOLOv8n.