@misc{indiciae833eca8ebc2b,
  title = {Learning When to Think: Shaping Adaptive Reasoning in R1-Style Models via Multi-Stage RL},
  author = {Songjun Tu and Jiahao Lin and Qichao Zhang and Xiangyu Tian and Linjing Li and Xiangyuan Lan and Dongbin Zhao},
  year = {2025},
  url = {https://arxiv.org/abs/2505.10832},
  note = {Source identifier: 2505.10832}
}