Paper Notes: Fine-Tuning Large Vision-Language Models as Decision-Making Agents via Reinforcement Learning