Transferability of the GCG Attack Across LLMs: Correlation with Attention Similarity and Transfer Rate Prediction.
Kazuki Takaki, Hirokazu Hasegawa, Yukiko Yamaguchi, Hajime Shimada
Browse the full COMPSAC paper archive.
Kazuki Takaki, Hirokazu Hasegawa, Yukiko Yamaguchi, Hajime Shimada
Browse the full COMPSAC paper archive.