SemStamp: A Semantic Watermark with Paraphrastic Robustness for Text Generation

arXiv:2310.03991v2 Announce Type: replace
Abstract: Existing watermarking algorithms are vulnerable to paraphrase attacks because of their token-level design. To address this issue, we propose SemStamp, a robust sentence-level semantic watermarking algorithm based on locality-sensitive hashing (LSH), which partitions the semantic space of sentences. The algorithm encodes and LSH-hashes a candidate sentence generated by an LLM, and conducts sentence-level rejection sampling until the sampled sentence falls in watermarked partitions in the semantic embedding space. A margin-based constraint is used to enhance …

abstract algorithm algorithms arxiv attacks cs.cl design hashes hashing issue lsh robust robustness semantic space text text generation the algorithm token type vulnerable watermark watermarking