VIG-RL: Learning to Search and Insert for Verified Image Grounding

Qinhan Yu, Jun Guang, Chong Chen +1

The paper introduces VIG-RL, a reinforcement‑learning based agent that dynamically decides when to retrieve, select, and insert authentic images into text responses, improving veri…