Publications

Refusal Is Layer-Distributed: A Causal Score for Where Safety Lives and an Adaptive Ablation That Uses It

Md Abrar Jahin, Md. Akramuddoula, Nawaz Talukder Sanglap, Md Rizwan Parvez, Jay Pujara, and Craig A Knoblock
Under review, 2026

BibTeX

@misc{jahin2026refusal,
  title = {{Refusal Is Layer-Distributed: A Causal Score for Where Safety Lives and an Adaptive Ablation That Uses It}},
  author = {Jahin, Md Abrar and Akramuddoula, Md. and Sanglap, Nawaz Talukder and Parvez, Md Rizwan and Pujara, Jay and Knoblock, Craig A},
  howpublished = {Under review},
  year = {2026}
}