Open-Set Visual Text Forensics via Sparse-Constraint Rectified Flow
2026-08-03 • Computer Vision and Pattern Recognition
Computer Vision and Pattern RecognitionArtificial Intelligence
AI summaryⓘ
The authors developed a new way to detect fake text edits in images created by advanced AI, which often tricks current detectors. Instead of trying to learn specific fake patterns, their method estimates how much change is needed to make the image look authentic locally, helping find tampered spots. They introduced a technique called Sparse-Constraint Rectified Flow (SC-RF) and used special tricks to handle limited data and keep important image details. Their method outperformed others on multiple tests and worked well even on new types of fake edits. They also showed that their approach can uncover weaknesses in existing detectors.
Generative AIforensic detectionflow matchinganomaly localizationself-supervised learningartifact injectionhigh-frequency forensic traceszero-shot learningimage tamperingtext editing detection
Authors
Jiangling Zhang, Shuxuan Gao, Zeyu Chen, Yichao Liu, Yu Zhou
Abstract
Rapidly evolving Generative AI enables sophisticated visual text manipulations that increasingly evade current forensic detectors. Existing discriminative models often overfit specific forgery patterns, limiting their generalization to unseen, open-set attacks. To address this challenge, we propose a generative detector that localizes tampering by estimating the local restoration cost required to align a query image with authentic visual-text statistics, rather than by learning forgery-specific decision boundaries. Specifically, we introduce Sparse-Constraint Rectified Flow (SC-RF), a detector-oriented adaptation of Flow Matching for spatially sparse anomaly localization. We further mitigate data scarcity via self-supervised Artifact Injection and preserve high-frequency forensic traces using a pixel-space Forensic-DiT. Extensive experiments on three benchmarks show that our method achieves state-of-the-art performance, surpassing the runner-up by 3.2 and 4.8 percentage points in F1 and IoU, respectively. In particular, the proposed detector demonstrates strong zero-shot performance on challenging unseen text editing patterns. We further provide an auxiliary stress-test analysis showing that local harmonization produced by our model can weaken the statistical cues relied upon by existing detectors, offering a complementary vulnerability-analysis perspective.