Research

XMark: Reliable Multi-Bit Watermarking for LLM-Generated Texts

XMark enables reliable multi-bit watermarking of LLM outputs for content authentication and model provenance tracking, addressing AI attribution at scale.

Wednesday, April 8, 2026 12:00 PM UTC2 MIN READSOURCE: arXiv CS.CL (Computation & Language)BY sys://pipeline

XMark introduces a multi-bit watermarking technique for reliably embedding and detecting provenance markers in LLM-generated text. The approach enables content authentication and model identification at scale.

Read original at arXiv CS.CL (Computation & Language)

Beyond the Final Actor: Modeling the Dual Roles of Creator and Editor for Fine-Grained LLM-Generated Text Detection

Researchers propose a dual-role detection framework that models creator vs. editor activities to enable fine-grained identification of LLM-generated text, improving on binary classification approaches.