Efficient Detection of LLM-generated Texts with a Bayesian Surrogate Model
This addresses the problem of inefficient detection of machine-generated text for applications needing real-time or low-cost monitoring, though it is incremental as it builds on existing methods like DetectGPT.
The paper tackles the inefficiency of detecting LLM-generated texts by proposing a Bayesian surrogate model that selects typical samples and interpolates scores, achieving significant performance gains with fewer queries; for example, it outperforms DetectGPT with 200 queries using only 2 or 3 queries on LLaMA-generated text.
The detection of machine-generated text, especially from large language models (LLMs), is crucial in preventing serious social problems resulting from their misuse. Some methods train dedicated detectors on specific datasets but fall short in generalizing to unseen test data, while other zero-shot ones often yield suboptimal performance. Although the recent DetectGPT has shown promising detection performance, it suffers from significant inefficiency issues, as detecting a single candidate requires querying the source LLM with hundreds of its perturbations. This paper aims to bridge this gap. Concretely, we propose to incorporate a Bayesian surrogate model, which allows us to select typical samples based on Bayesian uncertainty and interpolate scores from typical samples to other samples, to improve query efficiency. Empirical results demonstrate that our method significantly outperforms existing approaches under a low query budget. Notably, when detecting the text generated by LLaMA family models, our method with just 2 or 3 queries can outperform DetectGPT with 200 queries.