Adaptive Text Watermark for Large Language Models
This addresses the problem of misuse of AI-generated text for users needing to detect such content, though it appears incremental as it builds on existing watermarking methods.
The paper tackles the challenge of watermarking LLM-generated text by proposing an adaptive strategy that adds watermarks to high-entropy token distributions while leaving low-entropy ones untouched, and uses semantic embeddings to scale logits instead of fixed lists. The result is text with perplexity comparable to un-watermarked LLMs while maintaining robustness and security against attacks.
The advancement of Large Language Models (LLMs) has led to increasing concerns about the misuse of AI-generated text, and watermarking for LLM-generated text has emerged as a potential solution. However, it is challenging to generate high-quality watermarked text while maintaining strong security, robustness, and the ability to detect watermarks without prior knowledge of the prompt or model. This paper proposes an adaptive watermarking strategy to address this problem. To improve the text quality and maintain robustness, we adaptively add watermarking to token distributions with high entropy measured using an auxiliary model and keep the low entropy token distributions untouched. For the sake of security and to further minimize the watermark's impact on text quality, instead of using a fixed green/red list generated from a random secret key, which can be vulnerable to decryption and forgery, we adaptively scale up the output logits in proportion based on the semantic embedding of previously generated text using a well designed semantic mapping model. Our experiments involving various LLMs demonstrate that our approach achieves comparable robustness performance to existing watermark methods. Additionally, the text generated by our method has perplexity comparable to that of \emph{un-watermarked} LLMs while maintaining security even under various attacks.