Advancing Beyond Identification: Multi-bit Watermark for Large Language Models
Keywords: watermark, large language model, safety
TL;DR: We propose a method to watermark multibit information into language model outputs to tackle high-stake misuses.
Abstract: We propose a method to tackle misuses of large language models beyond the identification of machine-generated text. While existing methods focus on detection, some malicious misuses demand tracing the adversary user for counteracting them. To address this, we propose Multi-bit Watermark via Position Allocation, embedding traceable multi-bit information during language model generation. Leveraging the benefits of zero-bit watermarking, our method enables robust extraction of the watermark without any model access, embedding and extraction of long messages ($\geq$ 32-bit) without finetuning, and maintaining text quality, while allowing zero-bit detection all at the same time. Moreover, our watermark is relatively robust under strong attacks like interleaving human texts and paraphrasing.
Primary Area: societal considerations including fairness, safety, privacy
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2024/AuthorGuide.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors' identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 1040
Loading