WETBench: A Benchmark for Detecting Task-Specific Machine-Generated Text on Wikipedia

TL;DR

WETBench is a comprehensive benchmark for detecting machine-generated text on Wikipedia, focusing on realistic, task-specific scenarios across multiple languages and generators to improve detection reliability.

Contribution

Introduces WETBench, a multilingual, multi-generator, task-specific benchmark with new datasets for evaluating MGT detection in Wikipedia editing tasks.

Findings

Training-based detectors achieve 78% accuracy

Zero-shot detectors average 58% accuracy

Detectors struggle with realistic, task-specific MGT scenarios

Abstract

Given Wikipedia's role as a trusted source of high-quality, reliable content, concerns are growing about the proliferation of low-quality machine-generated text (MGT) produced by large language models (LLMs) on its platform. Reliable detection of MGT is therefore essential. However, existing work primarily evaluates MGT detectors on generic generation tasks rather than on tasks more commonly performed by Wikipedia editors. This misalignment can lead to poor generalisability when applied in real-world Wikipedia contexts. We introduce WETBench, a multilingual, multi-generator, and task-specific benchmark for MGT detection. We define three editing tasks, empirically grounded in Wikipedia editors' perceived use cases for LLM-assisted editing: Paragraph Writing, Summarisation, and Text Style Transfer, which we implement using two new datasets across three languages. For each writing task, we evaluate three prompts, generate MGT across multiple generators using the best-performing prompt, and benchmark diverse detectors. We find that, across settings, training-based detectors achieve an average accuracy of 78%, while zero-shot detectors average 58%. These results show that detectors struggle with MGT in realistic generation scenarios and underscore the importance of evaluating such models on diverse, task-specific data to assess their reliability in editor-driven contexts.