RADAR: Reasoning AI-Generated Image Detection for Semantic Fakes
As modern generative models advance rapidly, AI-generated images exhibit higher resolution and lifelike details. However, the generated images may not adhere to world knowledge and common sense, as there is no such awareness and supervision in the generative models. For instance, the generated image...
Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2025-07-01
|
| Series: | Technologies |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2227-7080/13/7/280 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1850071271505657856 |
|---|---|
| author | Haochen Wang Xuhui Liu Ziqian Lu Cilin Yan Xiaolong Jiang Runqi Wang Efstratios Gavves |
| author_facet | Haochen Wang Xuhui Liu Ziqian Lu Cilin Yan Xiaolong Jiang Runqi Wang Efstratios Gavves |
| author_sort | Haochen Wang |
| collection | DOAJ |
| description | As modern generative models advance rapidly, AI-generated images exhibit higher resolution and lifelike details. However, the generated images may not adhere to world knowledge and common sense, as there is no such awareness and supervision in the generative models. For instance, the generated images could feature a penguin walking in the desert or a man with three arms, scenarios that are highly unlikely to occur in real life. Current AI-generated image detection methods mainly focus on low-level features, such as detailed texture patterns and frequency domain inconsistency, which are specific to certain generative models, making it challenging to identify the above-mentioned general semantic fakes. In this work, (1) we propose a new task, reasoning AI-generated image detection, which focuses on identifying semantic fakes in generative images that violate world knowledge and common sense. (2) To benchmark the new task, we collect a new dataset Spot the Semantic Fake (STSF). STSF contains 358 images with clear semantic fakes generated by three different modern diffusion models and provides bounding boxes as well as text annotations to locate the fakes. (3) We propose RADAR, a reasoning AI-generated image detection assistor, to locate semantic fakes in the generative images and output corresponding text explanations. Specifically, RADAR contains a specialized multimodal LLM to process given images and detect semantic fakes. To improve the generalization ability, we further incorporate ChatGPT as an assistor to detect unrealistic components in grounded text descriptions. The experiments on the STSF dataset show that RADAR effectively detects semantic fakes in modern generative images. |
| format | Article |
| id | doaj-art-c7068815aa4e4c90837c6389ca90e331 |
| institution | DOAJ |
| issn | 2227-7080 |
| language | English |
| publishDate | 2025-07-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Technologies |
| spelling | doaj-art-c7068815aa4e4c90837c6389ca90e3312025-08-20T02:47:21ZengMDPI AGTechnologies2227-70802025-07-0113728010.3390/technologies13070280RADAR: Reasoning AI-Generated Image Detection for Semantic FakesHaochen Wang0Xuhui Liu1Ziqian Lu2Cilin Yan3Xiaolong Jiang4Runqi Wang5Efstratios Gavves6Informatics Institute, University van Amsterdam, 1098 XH Amsterdam, The NetherlandsSchool of Automation Science and Electrical Engineering, Beihang University, Beijing 100191, ChinaSchool of Computer Science and Technology (School of Artificial Intelligence), Zhejiang Sci-Tech University, Hangzhou 310018, ChinaSchool of Automation Science and Electrical Engineering, Beihang University, Beijing 100191, ChinaXiaohongshu Inc., Beijing 100029, ChinaSchool of Computer Science & Technology, Beijing Jiaotong University, Beijing 100091, ChinaInformatics Institute, University van Amsterdam, 1098 XH Amsterdam, The NetherlandsAs modern generative models advance rapidly, AI-generated images exhibit higher resolution and lifelike details. However, the generated images may not adhere to world knowledge and common sense, as there is no such awareness and supervision in the generative models. For instance, the generated images could feature a penguin walking in the desert or a man with three arms, scenarios that are highly unlikely to occur in real life. Current AI-generated image detection methods mainly focus on low-level features, such as detailed texture patterns and frequency domain inconsistency, which are specific to certain generative models, making it challenging to identify the above-mentioned general semantic fakes. In this work, (1) we propose a new task, reasoning AI-generated image detection, which focuses on identifying semantic fakes in generative images that violate world knowledge and common sense. (2) To benchmark the new task, we collect a new dataset Spot the Semantic Fake (STSF). STSF contains 358 images with clear semantic fakes generated by three different modern diffusion models and provides bounding boxes as well as text annotations to locate the fakes. (3) We propose RADAR, a reasoning AI-generated image detection assistor, to locate semantic fakes in the generative images and output corresponding text explanations. Specifically, RADAR contains a specialized multimodal LLM to process given images and detect semantic fakes. To improve the generalization ability, we further incorporate ChatGPT as an assistor to detect unrealistic components in grounded text descriptions. The experiments on the STSF dataset show that RADAR effectively detects semantic fakes in modern generative images.https://www.mdpi.com/2227-7080/13/7/280AI-generated detectionmultimodal large language modelssemantic-level fake detection |
| spellingShingle | Haochen Wang Xuhui Liu Ziqian Lu Cilin Yan Xiaolong Jiang Runqi Wang Efstratios Gavves RADAR: Reasoning AI-Generated Image Detection for Semantic Fakes Technologies AI-generated detection multimodal large language models semantic-level fake detection |
| title | RADAR: Reasoning AI-Generated Image Detection for Semantic Fakes |
| title_full | RADAR: Reasoning AI-Generated Image Detection for Semantic Fakes |
| title_fullStr | RADAR: Reasoning AI-Generated Image Detection for Semantic Fakes |
| title_full_unstemmed | RADAR: Reasoning AI-Generated Image Detection for Semantic Fakes |
| title_short | RADAR: Reasoning AI-Generated Image Detection for Semantic Fakes |
| title_sort | radar reasoning ai generated image detection for semantic fakes |
| topic | AI-generated detection multimodal large language models semantic-level fake detection |
| url | https://www.mdpi.com/2227-7080/13/7/280 |
| work_keys_str_mv | AT haochenwang radarreasoningaigeneratedimagedetectionforsemanticfakes AT xuhuiliu radarreasoningaigeneratedimagedetectionforsemanticfakes AT ziqianlu radarreasoningaigeneratedimagedetectionforsemanticfakes AT cilinyan radarreasoningaigeneratedimagedetectionforsemanticfakes AT xiaolongjiang radarreasoningaigeneratedimagedetectionforsemanticfakes AT runqiwang radarreasoningaigeneratedimagedetectionforsemanticfakes AT efstratiosgavves radarreasoningaigeneratedimagedetectionforsemanticfakes |