Ignacio Sanchez

SEP 3, 2026 · PREPRINT

Flip, Don't Shuffle: Watermarking LLMs at the Speed of Inference

arXiv

This is a preprint describing a novel computational method for LLM watermarking with theoretical proofs and benchmark results, but lacks peer review, clinical validation, or real-world deployment evidence needed for practice-changing claims.