Beyond Semantic Priors: Mitigating Optimization Collapse for Generalizable Visual Forensics
Institute of Information Engineering, Chinese Academy of Sciences（信息工程研究所，中国科学院）；School of Cyber Science and Technology, Beihang University（北京航空航天大学信息与科学学院）
Jipeng Liu, Haichao Shi, Siyu Xing, Rong Yin, Xiao-Yu Zhang
While Vision-Language Models (VLMs) like CLIP have emerged as a dominant paradigm for generalizable deepfake detection, a representational disconnect remains: their semantic-centric pre-training is ill-suited for capturing non-semantic artifacts inherent to hyper-realistic synthesis. In this work, we identify a failure mode termed Optimization Collapse, where detectors trained with Sharpness-Aware Minimization (SAM) degenerate to random guessing on non-semantic forgeries once the perturbation radius exceeds a narrow threshold. To theoretically formalize this collapse, we propose the Critical Optimization Radius (COR) to quantify the geometric stability of the optimization landscape, and leverage the Gradient Signal-to-Noise Ratio (GSNR) to measure generalization potential. We establish a theorem proving that COR increases monotonically with GSNR, thereby revealing that the geometric instability of SAM optimization originates from degraded intrinsic generalization potential. This result identifies the layer-wise attenuation of GSNR as the root cause of Optimization Collapse in detecting non-semantic forgeries. Although naively reducing perturbation radius yields stable convergence under SAM, it merely treats the symptom without mitigating the intrinsic generalization degradation, necessitating enhanced gradient fidelity. Building on this insight, we propose the Contrastive Regional Injection Transformer (CoRIT), which integrates a computationally efficient Contrastive Gradient Proxy (CGP) with three training-free strategies: Region Refinement Mask to suppress CGP variance, Regional Signal Injection to preserve CGP magnitude, and Hierarchical Representation Integration to attain more generalizable representations. Extensive experiments demonstrate that CoRIT mitigates optimization collapse and achieves state-of-the-art generalization across cross-domain and universal forgery benchmarks.
Seeing Before Reasoning: A Unified Framework for Generalizable and Explainable Fake Image Detection
Guangdong Provincial Key Laboratory of Intelligent Information Processing, Shenzhen Key Laboratory of Media Security, and SZU-AFS Joint Innovation Center for AI Technology（广东省智能信息处理重点实验室、深圳媒体安全重点实验室及深圳大学-AFS联合人工智能技术创新中心）；Tencent Youtu Lab（腾讯优图实验室）；Peking University（北京大学）；Sun Yat-Sen University（中山大学）
Deep Residual Injection for Full-Spectrum Forensic Signal Perception in Multimodal Large Language Models
National University of Singapore（新加坡国立大学）；Tsinghua University（清华大学）；University of Science and Technology of China（中国科学技术大学）；University of Electronic Science and Technology of China（电子科技大学）；University of California, Berkeley（加州大学伯克利分校）
Cultivating Forensic Reasoning for Generalizable Multimodal Manipulation Detection
School of Software Engineering, Xi’an Jiaotong University（西安交通大学软件工程学院）；School of Computer Science and Information Engineering, Hefei University of Technology（合肥工业大学计算机科学与信息工程学院）；CSIRO（澳大利亚联邦科学与工业研究组织）；Northwestern Polytechnical University（西北工业大学）；University of Macau（澳门大学）
When Detectors Forget Forensics: Blocking Semantic Shortcuts for Generalizable AI-Generated Image Detection
State Key Laboratory of Blockchain and Data Security, Zhejiang University（区块链与数据安全国家重点实验室，浙江大学）；National University of Singapore（新加坡国立大学）；Chongqing University of Posts and Telecommunications（重庆邮电大学）
Deepfake Forensics Adapter: A Dual-Stream Network for Generalizable Deepfake Detection
Shenzhen Technology University（深圳技术大学）；Singapore Institute of Technology（新加坡理工学院）；The University of Hong Kong（香港大学）；Nanyang Technological University（南洋理工大学）
Huawei Inc.（华为公司）；Institute of Automation, Chinese Academy of Sciences (CASIA)（中国科学院自动化研究所）；University of Chinese Academy of Sciences (UCAS)（中国科学院大学）；Alibaba DAMO Academy（阿里巴巴达摩院）
Pretrained vision-language models (VLMs) such as CLIP have shown impressive generalization capability in downstream vision tasks with appropriate text prompts. Instead of designing prompts manually, Context Optimization (CoOp) has been recently proposed to learn continuous prompts using taskspecific training data. Despite the performance improvements on downstream tasks, several studies have reported that CoOp suffers from the overfitting issue in two aspects: (i) the test accuracy on base classes first improves and then worsens during training;(ii) the test accuracy on novel classes keeps dec…
Breaking Latent Prior Bias in Detectors for Generalizable AIGC Image Detection
Guangdong Provincial Key Laboratory of Intelligent Information Processing, Shenzhen Key Laboratory of Media Security, and SZU-AFS Joint Innovation Center for AI Technology（广东智能信息处理省级重点实验室、深圳媒体安全重点实验室及SZU-AFS人工智能技术联合创新中心）；Nanchang University（南昌大学）；University of North Texas（北卡罗来纳州立大学）
Generalizable Synthetic Image Detection via Language-guided Contrastive Learning
School of Computer Science and Engineering, University of Electronic Science and Technology of China（电子科技大学计算机科学与工程学院）；State Key Laboratory of Internet of Things for Smart City（智慧城市物联网国家重点实验室）；Department of Computer and Information Science, Faculty of Science and Technology, University of Macau（澳门大学计算机与信息科学系）；Faculty of Applied Sciences, Macao Polytechnic University（澳门 polytechnic 大学应用科学系）
Mitigating Hallucinations in Large Vision-Language Models by Self-Injecting Hallucinations
Beijing Key Laboratory of Super Intelligent Security of Multi-Modal Information, CASIA（北京多模态信息超级智能安全重点实验室，中国科学院自动化所）；State Key Laboratory of Multimodal Artificial Intelligence Systems, CASIA（多模态人工智能系统国家重点实验室，中国科学院自动化所）；School of Artificial Intelligence, University of Chinese Academy of Sciences（中国科学院大学人工智能学院）；Hello Group（Hello集团）；Nanchang Hangkong University（南昌航空大学）；The University of Hong Kong（香港大学）；School of Information Science and Technology, ShanghaiTech University（上海科技大学信息科学与技术学院）
The Devil is in the Statistics: Mitigating and Exploiting Statistics Difference for Generalizable Semi-supervised Medical Image Segmentation
State Key Laboratory for Novel Software Technology, Nanjing University（南京大学计算机软件新技术国家重点实验室）；National Institute of Healthcare Data Science, Nanjing University（南京大学健康医疗大数据国家研究院）；School of Computer Science and Engineering, Southeast University（东南大学计算机科学与工程学院）；School of Data and Computer Science, Shandong Women’s University（山东女子学院数据与计算机科学学院）
ByteDance AI Lab（字节跳动人工智能实验室）；Harbin Institute of Technology（哈尔滨工业大学）
With the progress in AI-based facial forgery (i.e., deepfake), people are increasingly concerned about its abuse. Albeit effort has been made for training classification (also known as deepfake detection) models to recognize such forgeries, existing models suffer from poor generalization to unseen forgery technologies and high sensitivity to changes in image/video quality. In this paper, we advocate adversarial training for improving the generalization ability to both unseen facial forgeries and unseen image/video qualities. We believe training with samples that are adversarially crafted to at…
Exposing and Mitigating Temporal Attack in Deepfake Video Detection
Peking University（北京大学）；New York University（纽约大学）；Huzhou University（湖州大学）；Institute of Science Tokyo（东京科学研究院）；Guangdong Institute of Intelligence Science and Technology（广东省智能科学与技术研究院）
Generalizable Knowledge Distillation from Vision Foundation Models for Semantic Segmentation
School of Artificial Intelligence, Xidian University, China（西安电子科技大学人工智能学院）；Department of Information Engineering and Computer Science, University of Trento, Italy（特伦托大学信息工程与计算机科学系）；Department of Automation, Tsinghua University, China（清华大学自动化系）；School of Computer Science and Information Engineering, Hefei University of Technology, China（合肥工业大学计算机科学与信息工程学院）
NS-Net: Decoupling CLIP Semantic Information through NULL-Space for Generalizable AI-Generated Image Detection
Engineering Research Center of Digital Forensics, Ministry of Education, Nanjing University of Information Science and Technology（数字取证工程研究中心，教育部，南京信息科技大学）；University of Macau（澳门大学）
FairForensics: Seeing Expressions and Parsing Demographics via Vision-Language Modeling for Generalizable Fair Deepfake Detection
Shenzhen University（深圳大学）；Shandong Artificial Intelligence Institute（山东省人工智能研究院）；Qilu University of Technology (Shandong Academy of Sciences)（齐鲁工业大学（山东省科学院））；Tianjin University of Technology（天津理工大学）
Prototypical Progressive Alignment and Reweighting for Generalizable Semantic Segmentation
School of Computer Science and Cyber Engineering, Guangzhou University（计算机科学与网络安全学院，广州大学）；Guangdong Key Laboratory of Intelligent Information Processing, College of Electronics and Information Engineering, Shenzhen University（智能信息处理重点实验室，电子与信息工程学院，深圳大学）；School of Electronics and Communication Engineering, Guangzhou University（电子与通信工程学院，广州大学）；Univ Rennes, INSA Rennes, CNRS, IETR - UMR 6164（里昂大学，INSA里昂，CNRS，IETR - UMR 6164）；Guangdong Key Laboratory of Intelligent Information Processing, Shenzhen Key Laboratory of Advanced Machine Learning and Applications, Institute of Artificial Intelligence and Advanced Communication, College of Electronics and Information Engineering, Shenzhen University（智能信息处理重点实验室，先进机器学习与应用重点实验室，人工智能与高级通信研究所，电子与信息工程学院，深圳大学）；College of Mathematics and Statistics, Shenzhen University（数学与统计学院，深圳大学）
Generalizable Image Repair for Robust Visual Control
University of Florida AI Scholars Program（佛罗里达大学人工智能学者计划）；Trustworthy Engineered Autonomy (TEA) Lab（可信工程自主性实验室）；Department of Electrical and Computer Engineering（电气与计算机工程系）
MoreStyle: Relax Low-frequency Constraint of Fourier-based Image Reconstruction in Generalizable Medical Image Segmentation
School of Computer Science, Wuhan University（武汉大学计算机学院）；University of Louisville（路易斯维尔大学）；School of Computer Science, Central China Normal University（华中师范大学计算机学院）
Mitigating the Human-Robot Domain Discrepancy in Visual Pre-training for Robotic Manipulation
AI Thrust, The Hong Kong University of Science and Technology (Guangzhou)（香港科技大学（广州）AI thrust）；Sun Yat-sen University（中山大学）；The Hong Kong University of Science and Technology（香港科技大学）
