New watermarking method protects AI models from unauthorized copying

Original: Robust LLM Watermarking with Minimal Semantic Distortion for IP Protection

Source: arxiv.org ↗