Minimizing Mismatch Risk: A Prototype-Based Routing Framework for Zero-shot LLM-generated Text Detection

Zero-shot methods detect LLM-generated text by computing statistical signatures using a surrogate model. Existing approaches typically employ a fixed surrogate for all inputs regardless of the unknown source. We systematically examine this design and find that detection performance varies substantially depending on surrogate-source alignment. We observe that while no single surrogate achieves optimal performance universally, a well-matched surrogate typically exists within a diverse pool for any given input. This finding transforms robust detection into a routing problem: selecting the most appropriate surrogate for each input. We propose DetectRouter, a prototype-based framework that learns text-detector affinity through two-stage training. The first stage constructs discriminative prototypes from white-box models; the second generalizes to black-box sources by aligning geometric distances with observed detection scores. Experiments on EvoBench and MAGE benchmarks demonstrate consistent improvements across multiple detection criteria and model families.

Detecting AI-generated text is becoming increasingly important as large language models are widely used to produce articles, essays, comments, and other online content. Existing detection methods often rely on one fixed language model to judge whether a text was written by a human or generated by AI, but this can be unreliable because different AI systems leave different writing patterns. Our paper shows that choosing a better-matched detector for each input can greatly improve detection accuracy. We propose DetectRouter, a method that automatically selects the most suitable detector from a pool of candidate models based on the characteristics of the input text. Experiments on two large benchmarks show that this routing strategy improves robustness across different AI models and writing domains, making AI-generated text detection more practical for real-world use.