Overview
- Field: NLP
- Authors: Sondos Mahmoud Bsharat, Jiacheng Liu, Xiaohan Zhao
- Published: 2026-06-04
- arXiv: 2606.06481
- AI-text detectability is controlled not only by the proportion of AI-edited content, but also by the edit operation, domain, and cumulative revision history.
- Mixed-authorship intermediate versions are often harder to detect than both purely human and heavily AI-edited endpoints, exposing non-monotonic detection patterns — a phenomenon missed by existing benchmarks.
- OpAI-Bench provides a controlled testbed for analyzing when and how AI-assisted writing becomes detectable under realistic progressive editing scenarios.
Summary
As AI writing assistants become increasingly integrated into real-world drafting and revision workflows, many documents are no longer purely human-written or AI-generated, but instead result from progressive human-AI co-editing. However, existing AI-text detection benchmarks largely focus on final outputs and provide limited understanding of how AI authorship signals emerge, accumulate, or disappear throughout the revision process.
The authors introduce OpAI-Bench, an operation-guided benchmark for studying progressive human-to-AI text transformation across document, sentence, token, and span granularities. Starting from human-written documents, OpAI-Bench constructs nine sequentially revised versions for each sample under predefined AI coverage levels and five representative AI edit operations, covering four domains while preserving full authorship provenance at multiple granularities.
The benchmark supports comprehensive evaluation of 8 document-level, 7 sentence-level, and 2 fine-grained token/span detectors.
Key Findings
*Auto-collected on 2026-06-07*