Student Researcher, Peng Li’s Lab

• Investigated and mitigated hallucinations in Vision Language Models by correcting erroneous attention mechanisms, improving F1 score by 1% on the CHAIR benchmark compared to SOTA.