Beyond Easy Wins: A Text Hardness-Aware Benchmark for LLM-generated Text Detection
This addresses the need for more practical and equitable assessment of text detection systems, which is crucial for deployment in real-world scenarios, though it is incremental in improving evaluation methods.
The paper tackles the problem of evaluating AI text detectors by introducing a benchmark that prioritizes real-world reliability and stability, showing that current state-of-the-art zero-shot detection methods struggle with these aspects when tested with a hardness-aware humanification framework.
We present a novel evaluation paradigm for AI text detectors that prioritizes real-world and equitable assessment. Current approaches predominantly report conventional metrics like AUROC, overlooking that even modest false positive rates constitute a critical impediment to practical deployment of detection systems. Furthermore, real-world deployment necessitates predetermined threshold configuration, making detector stability (i.e. the maintenance of consistent performance across diverse domains and adversarial scenarios), a critical factor. These aspects have been largely ignored in previous research and benchmarks. Our benchmark, SHIELD, addresses these limitations by integrating both reliability and stability factors into a unified evaluation metric designed for practical assessment. Furthermore, we develop a post-hoc, model-agnostic humanification framework that modifies AI text to more closely resemble human authorship, incorporating a controllable hardness parameter. This hardness-aware approach effectively challenges current SOTA zero-shot detection methods in maintaining both reliability and stability. (Data and code: https://github.com/navid-aub/SHIELD-Benchmark)