Honglak Lee

Researchers tackle image understanding with the NICE challenge.

A novel approach improves graph representation learning using non-Euclidean spaces.

Combining 2D and 3D models boosts molecular property prediction accuracy.

A new method enhances caption quality for 3D objects.

RSCD enhances image quality in medical imaging, aiding in accurate diagnoses.

A new method, InsTa, enhances task selection in instruction tuning.

Small language models can enhance reasoning through self-generated critiques.

MASSW dataset simplifies scientific workflows through AI-generated summaries.

Transform discarded models into powerful new solutions through model merging.

Researchers connect data visualization with text creation for fresh insights.