Jul 2026
WaterMoE: Expert-Routing-based Watermarking for High Fidelity and Efficiency
This work proposes an LLM watermarking scheme \textit{WaterMoE} for the growingly popular Mixture-of-Experts (MoE) LLMs that achieves a fidelity performance close to the unwatermarked and consistently outperforms state-of-the-art watermarking methods on the benchmark.
Ze-Wen Sun, Qian Jiang, Si-Yuan Sheng et al.
· arXiv.org · 0 citations