The field of image forgery detection is moving towards more robust and generalizable methods, with a focus on adapting to diverse manipulation types and handling compressed images. Recent developments have introduced innovative frameworks that integrate multiple features and modules to improve detection accuracy and spatial localization. These advancements have led to state-of-the-art performance in various benchmark datasets and challenges. Notably, some papers have proposed novel solutions to address the challenges of deepfake detection on online social networks and remote sensing image forgery detection. Noteworthy papers include: Loupe, which proposes a lightweight yet effective framework for joint deepfake detection and localization, achieving state-of-the-art performance in the IJCAI 2025 Deepfake Detection and Localization Challenge. PLADA, which introduces a novel framework designed to tackle the lack of paired data and the ineffective use of compressed images in deepfake detection, demonstrating a remarkable balance in detection across 26 datasets. SFNet, which proposes a novel forgery detection framework that leverages spatial and frequency domain features to identify fake images in diverse remote sensing data, achieving an accuracy improvement of 4%-15.18% over state-of-the-art methods. M2SFormer, which proposes a novel Transformer encoder-based framework that unifies multi-frequency and multi-scale attentions to better capture diverse forgery artifacts, offering superior generalization in detecting and localizing forgeries across unseen domains.