SemanticShift: Robust Semantic Watermarking for Large Language Models
Proceedings of the AAAI Symposium Series · 18.05.2026 · 8 (1) · 455-464
DOI: 10.1609/aaaiss.v8i1.42577 ↗Kurz gefasst
Fachartikel in der Zeitschrift Proceedings of the AAAI Symposium Series, online erschienen am 18.05.2026. Verfasst von Minghao Li und 1 weiteren Autor:innen.
Abstract (Originalsprache)
Large language models (LLMs) have raised increasing concerns around misinformation and plagiarism. Watermarking—embedding identifiable signals into generated text—offers a promising approach for detection. Semantic watermarking enhances robustness by leveraging meaning rather than surface-level token patterns. However, existing semantic techniques often require model retraining or operate within constrained semantic spaces, limiting control over watermark strength, robustness, and cross-lingual generalizability. We introduce SemanticShift, a training-free and semantically grounded watermarking method that injects signals during generation by computing semantic shifts of candidate tokens relative to preceding context, guided by a secret key. SemanticShift uses pre-trained embedding models and is tunable via hyperparameters, offering strong resistance to paraphrasing and syntactic variation. Experiments demonstrate state-of-the-art detection performance, with ROC-AUC > 0.99 on original text and up to 0.96 under strong paraphrasing—outperforming all prior training-free approaches and rivaling training-based methods. Notably, SemanticShift achieves superior accuracy and robustness on models like OPT and LLaMA, showcasing its applicability and effectiveness.
Zitieren
APA
Li, M., & Tan, N. (2026). SemanticShift: Robust Semantic Watermarking for Large Language Models. Proceedings of the AAAI Symposium Series, 8(1), 455-464. https://doi.org/10.1609/aaaiss.v8i1.42577
Vancouver
Li M, Tan N. SemanticShift: Robust Semantic Watermarking for Large Language Models. Proceedings of the AAAI Symposium Series. 2026;8(1):455-464. doi:10.1609/aaaiss.v8i1.42577
BibTeX
@article{Li2026,
author = {Li, Minghao and Tan, Neset},
title = {SemanticShift: Robust Semantic Watermarking for Large Language Models},
journal = {Proceedings of the AAAI Symposium Series},
year = {2026},
volume = {8},
number = {1},
pages = {455--464},
doi = {10.1609/aaaiss.v8i1.42577}
}
EUVIMED