HammingMark improves semantic freedom in large language model watermarks
Beyond Semantic Narrowing: Robust and Efficient LLM Watermarking with Hamming Neighborhoods
Cryptography and SecurityArtificial Intelligence
Summary
Watermarks help track text generated by large language models, but existing methods can limit the way sentences are formed, making the writing less natural. The authors developed HammingMark, which uses a smart hashing technique to allow more natural sentence choices while still embedding a recognizable watermark. This approach reduces the need to try many sentence versions and keeps the meaning clear, even for tasks needing precise language. Tests showed HammingMark is more robust and efficient without hurting the quality of generated text.
What this means in practice
- •For content platform operators: Detect AI-generated text robustly without compromising natural language flow, reducing false positives from unnatural sentence constraints.
- •For automated summarization teams: Maintain high-quality, precise summaries while embedding watermarks to verify AI-generated content authenticity.
Authors
Zewen Sun, Tongyang Zhao, Liyao Xiang, Mingxuan Ma, Lingzhe Wang, Zhiyuan Li
Abstract
Semantic watermarking improves robustness against watermark removal attacks by embedding detectable signals into sentence-level representations. However, existing watermarking methods typically impose watermark-specific semantic preferences on generated sentences without explicitly accounting for the highly non-uniform and context-dependent semantic preference of LLM generation. When these two preferences are poorly aligned, many natural continuations become incompatible with the watermark, causing semantic narrowing: reduced semantic freedom, increased resampling cost, and potential degradation on tasks with strict semantic requirements. To alleviate this problem, we propose HammingMark, which uses the semantic hash of the preceding sentence as a dynamic center and accepts candidates whose hashes fall within its Hamming neighborhood. Defining watermark validity over a Hamming neighborhood in compact hash space retains a larger fraction of naturally likely semantic continuations. The coarse many-to-one hash mapping further allows diverse semantic realizations to remain watermark-valid. Experiments on C4 and BookSum show that HammingMark achieves strong robustness, high detectability, and near-unwatermarked generation quality, requiring only 2.2 sampled candidates per accepted sentence,a 72.8% reduction compared with the most sampling-efficient existing method. On more complex tasks with strict semantic constraints, HammingMark achieves the highest detection rates with the highest or tied-highest ROUGE-L scores, demonstrating its effectiveness in balancing watermark detectability and generation quality under constrained generation settings.