@misc{indiciaeb05f15ac86fe,
  title = {Knowledge Vector Weakening: Efficient Training-free Unlearning for Large Vision-Language Models},
  author = {Yejin Kim and Dongjun Hwang and Sungmin Cha and Junsuk Choe},
  year = {2026},
  url = {https://arxiv.org/abs/2601.21794},
  note = {Source identifier: 2601.21794}
}