MIRAGE: Towards AI-Generated Image Detection in the Wild

2025-08-17

AI Summary

Current AIGI detectors perform well under ideal laboratory conditions but exhibit severely limited generalization in realistic “in-the-wild” scenarios—characterized by heterogeneous generative sources, post-editing artifacts, and quality enhancement distortions. To address this gap, we introduce Mirage, the first high-challenge benchmark for in-the-wild AIGI detection, comprising internet-sourced and multi-model-synthesized images with complex, real-world noise. We further propose Mirage-R1, a vision-language model featuring a novel reflective reasoning mechanism and inference-time adaptive thinking strategies, enabling dynamic trade-offs between accuracy and latency. Training is optimized via supervised fine-tuning, reinforcement learning, and collaborative data generation from multiple expert generators. On Mirage, Mirage-R1 outperforms state-of-the-art methods by 5%; on public benchmarks, it achieves a 10% improvement, demonstrating significantly enhanced robustness for detecting AI-generated images in practical, unconstrained settings.

Technology Category

• Computer Vision: Large Vision Models
• Machine Learning: Large Multimodal Models (LMMs)
• Philosophy and Ethics of AI: Artificial General Intelligence

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

The spreading of AI-generated images (AIGI), driven by advances in generative AI, poses a significant threat to information security and public trust. Existing AIGI detectors, while effective against images in clean laboratory settings, fail to generalize to in-the-wild scenarios. These real-world images are noisy, varying from ``obviously fake" images to realistic ones derived from multiple generative models and further edited for quality control. We address in-the-wild AIGI detection in this paper. We introduce Mirage, a challenging benchmark designed to emulate the complexity of in-the-wild AIGI. Mirage is constructed from two sources: (1) a large corpus of Internet-sourced AIGI verified by human experts, and (2) a synthesized dataset created through the collaboration between multiple expert generators, closely simulating the realistic AIGI in the wild. Building on this benchmark, we propose Mirage-R1, a vision-language model with heuristic-to-analytic reasoning, a reflective reasoning mechanism for AIGI detection. Mirage-R1 is trained in two stages: a supervised-fine-tuning cold start, followed by a reinforcement learning stage. By further adopting an inference-time adaptive thinking strategy, Mirage-R1 is able to provide either a quick judgment or a more robust and accurate conclusion, effectively balancing inference speed and performance. Extensive experiments show that our model leads state-of-the-art detectors by 5% and 10% on Mirage and the public benchmark, respectively. The benchmark and code will be made publicly available.

Problem

• Detecting AI-generated images in real-world noisy environments
• Addressing generalization failure of existing detectors beyond lab settings
• Handling images from multiple generative models and post-editing

Innovation

• Mirage benchmark for in-the-wild AI-generated images
• Vision-language model with reflective reasoning mechanism
• Two-stage training with reinforcement learning optimization