LLM Text Watermarking

Reason about LLM text watermarking as a defender. Learn green-list / red-list approaches (Kirchenbauer et al.), SynthID-Text, the robustness-vs-quality trade-off, evaluation methodology (true-positive at low false-positive rate), the limit of short outputs, the paraphrase-attack resistance question, and deployment realities (cross-vendor, multi-model, opt-out).