TrueMoE: Dual-Routing Mixture of Discriminative Experts for Synthetic Image Detection

2025-09-19

AI Summary

Current synthetic image detection methods rely on a single, generic discriminative space, resulting in poor generalization and limited robustness against unseen generative patterns. To address this, we propose TrueMoE, a Mixture-of-Experts framework featuring dual routing mechanisms. TrueMoE partitions the discriminative space into specialized subspaces along two orthogonal axes: manifold structure and perceptual granularity. It introduces a synergistic routing strategy—sparse routing (granularity-aware) for coarse-level selection and dense routing (manifold-aware) for fine-grained adaptation—enabling input-dependent expert assignment. Each lightweight subspace is trained to capture distinct, complementary forgery cues. Extensive experiments across diverse generative models demonstrate that TrueMoE significantly outperforms state-of-the-art methods, achieving substantial gains in cross-architecture generalization and out-of-distribution robustness.

Technology Category

• Machine Learning: Mixture of Experts (MoE)
• Computer Vision: Generative Adversarial Networks (GANs) for Vision
• Search and Optimization: Learning to Search

Application Category

• Economics, Online Markets and Human Computation: Trust and reliance of crowd workers and data experts on GenAI
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

The rapid progress of generative models has made synthetic image detection an increasingly critical task. Most existing approaches attempt to construct a single, universal discriminative space to separate real from fake content. However, such unified spaces tend to be complex and brittle, often struggling to generalize to unseen generative patterns. In this work, we propose TrueMoE, a novel dual-routing Mixture-of-Discriminative-Experts framework that reformulates the detection task as a collaborative inference across multiple specialized and lightweight discriminative subspaces. At the core of TrueMoE is a Discriminative Expert Array (DEA) organized along complementary axes of manifold structure and perceptual granularity, enabling diverse forgery cues to be captured across subspaces. A dual-routing mechanism, comprising a granularity-aware sparse router and a manifold-aware dense router, adaptively assigns input images to the most relevant experts. Extensive experiments across a wide spectrum of generative models demonstrate that TrueMoE achieves superior generalization and robustness.

Problem

• Detecting synthetic images from diverse generative models
• Overcoming limited generalization in unified detection spaces
• Capturing diverse forgery cues across specialized subspaces

Innovation

• Dual-routing mixture of experts framework
• Specialized lightweight discriminative subspaces collaboration
• Granularity-aware and manifold-aware adaptive routing mechanism