GlobalForge: Towards Robust AI-Generated Image Detection

2026-07-16

AI Summary

Current AI-generated image detectors suffer significant performance degradation due to their reliance on local artifacts that are easily disrupted by common transmission degradations such as JPEG compression and blurring. To address this limitation, this work proposes GlobalForge, a novel framework that suppresses fragile local cues through a Local Information Bottleneck (LIB) and leverages a Global Structure Reasoning (GSR) module to guide the model toward degradation-robust global semantic structures. Furthermore, we introduce a degradation-aware contrastive structural loss and construct RealDeg-Bench, a comprehensive evaluation benchmark encompassing seven types of single and composite degradations. Extensive experiments demonstrate that our approach achieves an average balanced accuracy (BAcc) improvement of 5.89% across eight real-world benchmarks and significantly outperforms existing methods on RealDeg-Bench.

Abstract

AI-generated image (AIGI) detectors achieve strong accuracy on clean benchmarks, but their performance drops sharply after images are propagated through real-world channels. We trace this fragility to what these detectors actually learn: they overfit to local artifacts left by generators in small spatial neighborhoods, which are easily destroyed by common propagation degradations such as JPEG compression and blur. Instead, we shift the discriminative cue from fragile local artifacts to more robust global structure. Building on this, we propose GlobalForge, a framework with two complementary modules. The Local Information Bottleneck (LIB) suppresses local components to block shortcut learning, while the Global Structural Reasoning (GSR) module forces every token to gather evidence from distant regions. Both modules are trained jointly under a contrastive structural loss based on degradation that keeps the resulting features stable under degradation. To support fine-grained robustness evaluation, we further introduce RealDeg-Bench, covering 7 common degradation operators and multi-step compound chains. GlobalForge improves average BAcc on 8 in-the-wild benchmark groups by $\mathbf{5.89\%}$ over the previous state-of-the-art, and is clearly ahead of representative baselines on RealDeg-Bench under both single and compound degradations. Code is available at https://anonymous.4open.science/r/GlobalForge-BE0F/.

Problem

• AI-generated image detection
• robustness
• image degradation
• real-world propagation
• detector fragility

Innovation

• Global Structural Reasoning
• Local Information Bottleneck
• Robust AIGI Detection
• Contrastive Structural Loss
• RealDeg-Bench