A black-box watermarking method using paired positive and negative backdoor triggers and shadow-LoRA training to keep detection working under LoRA addition, negation, and multi-LoRA merging.
A watermark for large language models
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CR 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
LoRAGuard: An Effective Black-box Watermarking Approach for LoRAs
A black-box watermarking method using paired positive and negative backdoor triggers and shadow-LoRA training to keep detection working under LoRA addition, negation, and multi-LoRA merging.