Combining acrostic, sensorimotor, and red-green watermark features in an ensemble improves LLM watermark detection after paraphrasing from 49% to 95%.
Defending Against Neural Fake News
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Ensemble Watermarks for Large Language Models
Combining acrostic, sensorimotor, and red-green watermark features in an ensemble improves LLM watermark detection after paraphrasing from 49% to 95%.