ForensicsTok: Forensics-Guided Tokenized Modeling for Image Tampering Localization
This work addresses the problem of precise image tampering localization for forensic applications, offering a unified architecture that improves over existing MLLM-based methods and competes with specialized forensic models.
ForensicsTok reformulates image tampering localization as an autoregressive sequence generation task, directly generating spatially grounded token sequences without intermediary supervision. It achieves substantial improvements over MLLM-based baselines and slight improvements over strong forensic expert baselines on six benchmarks, with stronger robustness to perturbations.
Multi-modal Large Language Models (MLLMs) offer powerful reasoning for forensic tasks, yet existing approaches utilizing exogenous segmentation decoders often suffer from suboptimal localization. The reliance on stitched pipelines introduces information bottlenecks during backpropagation, which dilutes spatial signals and is limited by semantic priors of the segmentor. To address these limitations, we propose ForensicsTok, which reformulates image manipulation localization as an autoregressive sequence generation task. ForensicsTok directly generates spatially grounded token sequences, enabling precise mask prediction without intermediary supervision. Specifically, we introduce a Token Splatting Decoder (TSD) to map tokens to binary masks via codebook-aware code smoothing, which mitigates sharp gradients from deterministic detokenizers. Furthermore, to capture diverse tampering clues, we propose a Hierarchical Expert Fusion (HEF) module that injects multi-scale features from a forensic expert model. This unified architecture effectively compensates for the lack of forensic priors in standard MLLMs. Extensive experiments on six benchmarks show that ForensicsTok substantially improves over existing MLLM-based baselines and slightly improves over strong forensic expert baselines, while exhibiting stronger robustness to perturbations.