To Think or Not To Think: A Study of Thinking in Rule-Based Visual Reinforcement Fine-Tuning | Read Paper on Bytez