A Watermark for Large Language Models
Kirchenbauer et al., 2023, A Watermark for Large Language Models (Proceedings of the 40th International Conference on Machine Learning (ICML 2023), PMLR 202)
Subscriber deck — free during early access
1 cards in “8 Related Work” · whole deck
8 Related Work
- Why do Kirchenbauer et al. prefer watermarking to post-hoc detection of machine text?Post-hoc detectors lose ground as models improve (GPT-2 detectors struggle with GPT-3), are vulnerable to adversarial attacks, and risk falsely flagging human text, especially from non-native speakers or writers using assistance tools.
- Kirchenbauer et al., 2023, A Watermark for Large Language Models (Proceedings of the 40th International Conference on Machine Learning (ICML 2023), PMLR 202), §8
Source
- John Kirchenbauer, Jonas Geiping, Yuxin Wen, Jonathan Katz, Ian Miers and Tom Goldstein. 2023. A Watermark for Large Language Models. In Proceedings of the 40th International Conference on Machine Learning (ICML 2023), PMLR 202, pp. 17061–17084. https://proceedings.mlr.press/v202/kirchenbauer23a.html