WETBench: A Benchmark for Detecting Task-Specific Machine-Generated Text on Wikipedia
This addresses the issue of low-quality machine-generated content on Wikipedia for editors and users, but it is incremental as it focuses on benchmarking rather than proposing new detection methods.
The paper tackles the problem of detecting machine-generated text on Wikipedia by introducing WETBench, a task-specific benchmark, and finds that detectors achieve only 78% accuracy for training-based methods and 58% for zero-shot methods in realistic scenarios.
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.