WaterMax is a groundbreaking watermarking scheme for Large Language Models (LLMs) that balances detectability and quality. Unlike other methods, WaterMax does not require modifications to the LLM itself, ensuring robustness and complexity. The scheme improves text quality and watermark detectability by working with token sequences rather than individual tokens. Theoretical models validate its performance under attack scenarios. WaterMax offers a new approach to watermarking LLM-generated text, addressing key challenges in the field.
Til et andet sprog
fra kildeindhold
arxiv.org
Vigtigste indsigter udtrukket fra
by Eva Giboulot... kl. arxiv.org 03-11-2024
https://arxiv.org/pdf/2403.04808.pdfDybere Forespørgsler