Zongxian Yang

Zongxian Yang

Research Assistant at the HANI Lab, working on LLM reasoning, medical AI, and multimodal reasoning.

Med-REFL: Medical Reasoning Enhancement via Self-Corrected Fine-grained Reflection

Published in arXiv preprint, 2025

Med-REFL is a project that I led from idea formation to experiments and writing. The core goal is to improve medical reasoning reliability by building fine-grained reflection and self-correction signals into the reasoning and training process.