Xiaomeng Li

Forget blindly trusting AI tools: this agent learns which ones are actually reliable for different medical image analysis tasks.

Overcome the "forgetting" problem in source-free domain adaptation by explicitly preserving confident predictions, leading to state-of-the-art fundus image diagnosis.

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