SIRA is a black-box paraphrase attack that masks high-self-information tokens and fills the gaps with an LLM, achieving near-100% watermark removal on seven schemes.
Claude 3.5, 2024
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Revealing Weaknesses in Text Watermarking Through Self-Information Rewrite Attacks
SIRA is a black-box paraphrase attack that masks high-self-information tokens and fills the gaps with an LLM, achieving near-100% watermark removal on seven schemes.