@misc{indiciaee878fddadcd7,
  title = {Attention Degradation, Function Token Anchoring, and the Limits of Attention-Based Intervention in Large Language Models},
  author = {Sagar Dangal and Manoj Shakya},
  year = {2026},
  url = {https://arxiv.org/abs/2607.20524},
  note = {Source identifier: 2607.20524}
}