PhD student, University of Edinburgh
2 papers at NeurIPS 2025
We propose Magical, an asymmetric LoRA architecture to meet the requirement for semantic fidelity and diverse lay-style generation in medical lay language generation task.
MedAgentBoard finds multi-agent useful for some medical tasks, but single LLMs or conventional methods are often better, especially for VQA/EHR. Task-specific AI choice is key.