Toward Generalizable Forgery Detection and Reasoning
- 26 March 2026
- journal article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Image Processing
- Vol. 35 (10577149) , 3395-3410
Abstract
Accurate and interpretable detection of AI-generated images is essential for mitigating risks associated with AI misuse. However, the substantial domain gap among generative models makes it challenging to develop a generalizable forgery detection model. Moreover, since every pixel in an AI-generated image is synthesized, traditional saliency-based forgery explanation methods are not well suited for this task. To address these challenges, we formulate detection and explanation as a unified Forgery Detection and Reasoning task (FDR-Task)1, leveraging Multi-Modal Large Language Models (MLLMs) to provide accurate detection through reliable reasoning over forgery attributes. To facilitate this task, we introduce the Multi-Modal Forgery Reasoning dataset (MMFR-Dataset), a large-scale dataset containing 120K images across 10 generative models, with 378K reasoning annotations on forgery attributes, enabling comprehensive evaluation of the FDR-Task. Furthermore, we propose FakeReasoning, a forgery detection and reasoning framework with three key components: 1) a dual-branch visual encoder that integrates CLIP and DINO to capture both high-level semantics and low-level artifacts; 2) a Forgery-Aware Feature Fusion Module that leverages DINO’s attention attention maps and cross-attention mechanisms to guide MLLMs toward forgery-related clues; 3) a Classification Probability Mapper that couples language modeling and forgery detection, enhancing overall performance. Experiments across multiple generative models demonstrate that FakeReasoning not only achieves robust generalization but also outperforms state-of-the-art methods on both detection and reasoning tasks.Keywords
All Related Versions
Funding Information
- National Natural Science Foundation of China (62406171)
- National Natural Science Foundation of China (62225601)
- National Natural Science Foundation of China (U23B2052)
- National Natural Science Foundation of China (62406172)
- National Natural Science Foundation of China (L242025)
- Fundamental Research Funds for the Beijing University of Posts and Telecommunications (2025AI4S15)
- China Postdoctoral Science Foundation (2023M741964)
- the Postdoctoral Fellowship Program of CPSF (GZC20240841)
This publication has 44 references indexed in Scilit:
- Common Sense Reasoning for Deepfake Detection2024
- Eyes Wide Shut? Exploring the Visual Shortcomings of Multimodal LLMs2024
- BioInstruct: instruction tuning of large language models for biomedical natural language processingJournal of the American Medical Informatics Association, 2024
- Hierarchical Fine-Grained Image Forgery Detection and Localization2023
- Towards Universal Fake Image Detectors that Generalize Across Generative Models2023
- TruFor: Leveraging All-Round Clues for Trustworthy Image Forgery Detection and Localization2023
- GRES: Generalized Referring Expression Segmentation2023
- RePaint: Inpainting using Denoising Diffusion Probabilistic Models2022
- CNN-Generated Images Are Surprisingly Easy to Spot… for Now2020
- ImageNet Large Scale Visual Recognition ChallengeInternational Journal of Computer Vision, 2015