@misc{indiciae3800722df53a,
  title = {Shifting the Gradient: Understanding How Defensive Training Methods Protect Language Model Integrity},
  author = {Satchel Grant and Victor Gillioz and Jake Ward and Thomas McGrath},
  year = {2026},
  url = {https://arxiv.org/abs/2604.16423},
  note = {Source identifier: 2604.16423}
}