EvoGuard: An Extensible Agentic RL-based Framework for Practical and Evolving AI-Generated Image Detection

2026-03-18

AI Summary

This work addresses the growing risk of misinformation posed by the proliferation of AI-generated images (AIGIs), for which existing detection methods suffer from limited generalizability, scalability, and high annotation costs. To overcome these challenges, the authors propose an agent-based reinforcement learning framework that adaptively orchestrates off-the-shelf multimodal and non-multimodal detectors through a capability-aware dynamic coordination mechanism. The framework features a training-free, plug-in architecture that requires only binary labels for optimization and supports seamless integration of new detectors without retraining. It integrates multimodal large language models with agent-driven planning and self-reflection capabilities, trained via the GRPO algorithm. Evaluated across multiple benchmarks, the approach achieves state-of-the-art accuracy, effectively mitigates positive-negative sample bias, and enables continual performance improvement without additional training.

Technology Category

• Natural Language Processing: Fact-Checking / Misinformation Detection (NLP Focus)
• Computer Vision: Multi-modal Vision
• Multiagent Systems: Adversarial Agents

Application Category

• Search and Retrieval-Augmented AI: Agentic search
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

The rapid proliferation of AI-Generated Images (AIGIs) has introduced severe risks of misinformation, making AIGI detection a critical yet challenging task. While traditional detection paradigms mainly rely on low-level features, recent research increasingly focuses on leveraging the general understanding ability of Multimodal Large Language Models (MLLMs) to achieve better generalization, but still suffer from limited extensibility and expensive training data annotations. To better address complex and dynamic real-world environments, we propose EvoGuard, a novel agentic framework for AIGI detection. It encapsulates various state-of-the-art (SOTA) off-the-shelf MLLM and non-MLLM detectors as callable tools, and coordinates them through a capability-aware dynamic orchestration mechanism. Empowered by the agent's capacities for autonomous planning and reflection, it intelligently selects suitable tools for given samples, reflects intermediate results, and decides the next action, reaching a final conclusion through multi-turn invocation and reasoning. This design effectively exploits the complementary strengths among heterogeneous detectors, transcending the limits of any single model. Furthermore, optimized by a GRPO-based Agentic Reinforcement Learning algorithm using only low-cost binary labels, it eliminates the reliance on fine-grained annotations. Extensive experiments demonstrate that EvoGuard achieves SOTA accuracy while mitigating the bias between positive and negative samples. More importantly, it allows the plug-and-play integration of new detectors to boost overall performance in a train-free manner, offering a highly practical, long-term solution to ever-evolving AIGI threats. Source code will be publicly available upon acceptance.

Problem

• AI-Generated Image Detection
• Multimodal Large Language Models
• Extensibility
• Dynamic Environments
• Misinformation

Innovation

• Agentic Reinforcement Learning
• Multimodal Large Language Models
• Dynamic Orchestration
• AI-Generated Image Detection
• Plug-and-Play Integration