Detecting Post-generation Edits to Watermarked LLM Outputs via Combinatorial Watermarking

Open in new window