TY  - RPRT
TI  - Linearly Decoding Refused Knowledge in Aligned Language Models
AU  - Aryan Shrivastava
AU  - Ari Holtzman
PY  - 2025
UR  - https://arxiv.org/abs/2507.00239
ID  - 2507.00239
ER  -