Search papers, labs, and topics across Lattice.
This paper introduces ToolDF, a novel framework for detecting mixed-authenticity audio deepfakes that integrates tool-use reasoning with an audio large language model. By adaptively analyzing audio scenes and utilizing domain-specific experts, ToolDF not only identifies manipulated audio but also localizes the evidence supporting its decisions. Experimental results demonstrate that ToolDF outperforms existing methods, achieving significant macro-F1 score improvements while offering interpretable insights into the detection process.
ToolDF achieves a remarkable 14.39-point boost in macro-F1 scores over traditional methods while providing interpretable evidence for mixed-authenticity audio deepfake detection.
Audio deepfake detection is commonly formulated as clip-level binary classification of single-domain audio. However, real-world manipulated audio can exhibit mixed authenticity, where genuine and manipulated cues coexist across temporal transitions, overlapping sources, or both. This setting requires not only detecting manipulated audio but also localizing the components that provide evidence for the decision. We propose ToolDF, a tool-integrated reasoning framework for mixed-authenticity audio deepfake detection. ToolDF employs an audio large language model as an orchestrator trained with supervised tool-use trajectories. It adaptively analyzes the audio scene, selectively performs source separation, routes components to domain-specific experts, and aggregates their evidence into an interpretable verdict. We further introduce a mixed-authenticity ADD benchmark covering temporal transitions, acoustic overlaps, and hybrid mixtures. Experimental results show that ToolDF achieves the best overall performance on composite-type detection, achieving macro-F1 gains of 3.72 and 14.39 points over the strongest monolithic baseline and a fixed pipeline, respectively, while providing interpretable evidence localized to temporal regions and acoustic sources. Our source code and dataset are publicly available online.
