SEAL embeds a secret matrix between LoRA's low-rank factors, then decomposes and hides it in the released weights, but its ownership verification can be gamed by claiming the identity matrix as the passport.
Refit: A unified watermark removal framework for deep learning systems with limited data
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
SEAL: Entangled White-box Watermarks on Low-Rank Adaptation
SEAL embeds a secret matrix between LoRA's low-rank factors, then decomposes and hides it in the released weights, but its ownership verification can be gamed by claiming the identity matrix as the passport.