Search papers, labs, and topics across Lattice.
This paper addresses the challenges of continual learning in AI-generated image detection, particularly the issues of representation and decision degradation that lead to catastrophic forgetting. The authors introduce DECODE, a framework that employs Subspace Diversity Regularization and Closed-Form Decision Alignment to maintain both feature representations and decision boundaries across multiple generative domains. Experimental results demonstrate that DECODE achieves an impressive average accuracy of 99.36% with minimal forgetting, while also generalizing effectively to previously unseen generators.
Even with stable feature representations, AI detectors can still fail due to drifting decision boundaries, leading to a new failure mode called Dual Degradation.
As generative models continue to evolve, AI-generated image detectors must incrementally adapt to emerging generative domains while preserving knowledge acquired from previous ones. This continual learning setting is particularly challenging because forensic traces are often subtle and generator-specific, making detectors highly vulnerable to catastrophic forgetting. Existing methods primarily address this problem by stabilizing feature representations, implicitly treating forgetting as a representation-level issue. In this paper, we show that this perspective is incomplete. We demonstrate that even when feature representations remain discriminative, the decision boundary can progressively drift as the classification head is continually optimized on new domains. These two effects jointly give rise to a compound failure mode, termed Dual Degradation. To overcome this challenge, we propose DECODE, a decoupled continual detection framework that jointly mitigates representation- and decision-level forgetting. Specifically, we introduce Subspace Diversity Regularization (SDR) to preserve diverse forensic representations and Closed-Form Decision Alignment (CDA) to recalibrate the shared classification head after each adapter merge without manual hyperparameter tuning. Extensive experiments on 19 generative domains show that DECODE achieves an average accuracy of 99.36% with only 0.39% forgetting, while further generalizing to 11 unseen generators with 95.36% accuracy.
