Reinforcing VLMs to Use Tools for Detailed Visual Reasoning Under Resource Constraints

Open in new window