{"id":"W4416223010","doi":"10.48550/arxiv.2511.08985","title":"DeepTracer: Tracing Stolen Model via Deep Coupled Watermarks","year":2025,"lang":"en","type":"preprint","venue":"ArXiv.org","topic":"Adversarial Robustness in Machine Learning","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Instituto de Ciencias del Mar y Limnología, Universidad Nacional Autónoma de México; Indian Council of Medical Research; Institute for Catastrophic Loss Reduction","keywords":"Digital watermarking; Watermark; Task (project management); Key (lock); Reliability (semiconductor); Tracing","routes":{"ca_aff":false,"ca_fund":true,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":["metaepi_narrow","research_integrity"],"consensus_categories":[],"category_scores_codex":[0.0008151854,0.0007089005,0.0008119418,0.0003437426,0.0004170293,0.0002775123,0.003678104,0.0006869699,0.00004605238],"category_scores_gemma":[0.000193304,0.0007274118,0.0003533704,0.0003529346,0.0001028651,0.000577363,0.005280903,0.002608507,0.000112408],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0003185535,"about_ca_system_score_gemma":0.000407681,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0002302481,"about_ca_topic_score_gemma":0.00005433715,"domain_scores_codex":[0.9958774,0.000258768,0.0007821397,0.001636805,0.0006035164,0.0008413152],"domain_scores_gemma":[0.9966352,0.0002491888,0.0004703738,0.002216057,0.000220887,0.0002083214],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.00001768036,0.00005796815,0.01672887,0.0001939532,0.0001116762,0.00005888562,0.00176346,0.9690686,0.0002326254,0.0007448682,0.00006791149,0.01095353],"study_design_scores_gemma":[0.0004436149,0.00001767855,0.004668487,0.0002406934,0.00007970438,0.000009982798,0.00002296868,0.9898744,0.0003077562,0.003492669,0.0001236002,0.0007184298],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.1466612,0.000316345,0.8469154,0.0008869489,0.001830096,0.0004471345,0.000004855389,0.0006938656,0.002244135],"genre_scores_gemma":[0.9006975,0.00005190452,0.09676772,0.0005034658,0.0002922206,0.00008547848,0.00004235591,0.00005077015,0.001508599],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.7540363,"threshold_uncertainty_score":0.9996925,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02529531175863155,"score_gpt":0.2790813172186406,"score_spread":0.2537860054600091,"validation_status":"score_only:v0-immature-baseline","note":"Baseline scores from an immature model (maturity gate not passed). Scores rank; they never assert a category."}}