@misc{indiciae2fb4ee6afe56,
  title = {Depth Improves Training-Time Scaling for Difficult Tasks in Deep Linear Networks},
  author = {Taishi Watanabe and Ryo Karakida and Jun-nosuke Teramae},
  year = {2026},
  url = {https://arxiv.org/abs/2610.35514},
  note = {Source identifier: 2610.35514}
}