{"ID":5935835,"CreatedAt":"2026-07-07T01:22:02.77346169Z","UpdatedAt":"2026-07-07T02:10:06.972658124Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2607.03134","arxiv_id":"2607.03134","title":"Open-Set Source Tracing as Compositional Factors via Structured Prototypes","abstract":"Recent research expands beyond binary anti-spoofing with the emergence of Source Tracing, the task of identifying the specific generative origins of synthetic speech. However, current research often equates a \"source\" with its generative architecture. We propose redefining a source as a compositional tuple of Architecture, Training Data, and other training factors affecting the generated speech. We propose a framework using Structured Orthonormal Prototypes to minimize class overlap and intra-class variance. Our Subspace Partitioning strategy splits the embedding into architecture and data subspaces, while a residual subspace captures stochastic variability, enabling \"compositional generalization\" for novel factor combinations. This approach improves performance for partially seen sources and maintains robustness in fully open-set scenarios. MLAAD evaluations for Few-Shot open-set Identification show our approach significantly outperforms angular-margin baselines.","short_abstract":"Recent research expands beyond binary anti-spoofing with the emergence of Source Tracing, the task of identifying the specific generative origins of synthetic speech. However, current research often equates a \"source\" with its generative architecture. We propose redefining a source as a compositional tuple of Architect...","url_abs":"https://arxiv.org/abs/2607.03134","url_pdf":"https://arxiv.org/pdf/2607.03134v1","authors":"[\"Santiago Rubio\",\"Antonio Almudévar\",\"Antonio Miguel\",\"Eduardo Lleida\",\"Alfonso Ortega\"]","published":"2026-07-03T09:23:36Z","proceeding":"eess.AS","tasks":"[\"eess.AS\",\"cs.LG\"]","methods":"[]","has_code":false}