Siqi Liu

Resolving semantic conflicts between synonymous prompts and across categories dramatically improves the stability and accuracy of open-vocabulary semantic segmentation, all without requiring any additional training.

Achieve SOTA in cross-domain graph few-shot learning by adaptively aligning diverse graph features into a shared semantic space.

Zooming out to mixed-magnification views in computational pathology unlocks cancer-specific improvements in biomarker prediction, suggesting that spatial context is more crucial than previously thought.