Jul 2026
VIG-RL: Learning to Search and Insert for Verified Image Grounding
VIG-RL is an autonomous agentic framework that formulates the search-selection-insertion workflow as an active decision-making process, guided by a composite reward system that holistically evaluates the agent's step-by-step tool execution and final multimodal alignment.
Qinhan Yu, Jun Guang, Chong Chen et al.
· arXiv.org · 0 citations