Sparse Watermarking in LLMs with Enhanced Text Quality

The dataset used in the paper is not explicitly described, but it is mentioned that the authors used the ELI5, FinanceQA, MultiNews, and QMSum datasets.

BibTex: