一作ACL 2026Findings· Findings of ACL
The Mark Fades: Adaptive Evolutionary Paraphrase-based Attack against LLM Watermarks
Yusheng Zhao · Jian Zhao · Tianle Zhang · Feng Wei · Xuelong Li
摘要
While LLM watermarking is essential for machine- generated content identification, existing paraphrase-based attacks struggle to balance watermark removal efficacy with text quality. We propose TSAPA, a training-free evolutionary framework that models watermark removal as a constrained multi-obje...