Enhancing Vision Language Models Through Preference Optimization and Hallucination Suppression

Explore preference optimization and reward models for reducing hallucinations and improving zero-shot reasoning in VLMs.