AGENT-X: Adaptive Guideline-based Expert Network for Threshold-free AI-generated teXt detection

2025-05-21

AI Summary

Current AI-generated text detection methods rely heavily on large-scale labeled datasets and manually set thresholds, resulting in poor interpretability, weak zero-shot generalization, and limited cross-domain adaptability. To address these limitations, we propose the first zero-shot multi-agent detection framework that eliminates threshold-based decision-making. Our approach integrates rhetoric and systemic functional linguistics to construct a three-dimensional, interpretable detection guideline. It employs semantic-guided confidence calibration, confidence-aware meta-aggregation, and an adaptive Mixture-of-Agents router to enable explicit reasoning and dynamic guideline routing. Evaluated on diverse, multi-source, cross-domain benchmarks, our method achieves state-of-the-art performance in accuracy, interpretability, and zero-shot generalization—significantly outperforming existing supervised and zero-shot baselines.

Technology Category

• Natural Language Processing: Safety and Robustness
• Multiagent Systems: Adversarial Agents
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Existing AI-generated text detection methods heavily depend on large annotated datasets and external threshold tuning, restricting interpretability, adaptability, and zero-shot effectiveness. To address these limitations, we propose AGENT-X, a zero-shot multi-agent framework informed by classical rhetoric and systemic functional linguistics. Specifically, we organize detection guidelines into semantic, stylistic, and structural dimensions, each independently evaluated by specialized linguistic agents that provide explicit reasoning and robust calibrated confidence via semantic steering. A meta agent integrates these assessments through confidence-aware aggregation, enabling threshold-free, interpretable classification. Additionally, an adaptive Mixture-of-Agent router dynamically selects guidelines based on inferred textual characteristics. Experiments on diverse datasets demonstrate that AGENT-X substantially surpasses state-of-the-art supervised and zero-shot approaches in accuracy, interpretability, and generalization.

Problem

• Eliminates need for large annotated datasets and threshold tuning
• Enhances interpretability and adaptability in AI-text detection
• Improves zero-shot effectiveness across diverse textual characteristics

Innovation

• Zero-shot multi-agent framework for text detection
• Semantic steering for robust confidence calibration
• Adaptive Mixture-of-Agent router for dynamic guideline selection