{"id":"W4417035276","doi":"10.48550/arxiv.2512.04008","title":"Efficient Public Verification of Private ML via Regularization","year":2025,"lang":"","type":"preprint","venue":"ArXiv.org","topic":"Privacy-Preserving Technologies in Data","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Natural Sciences and Engineering Research Council of Canada; Defense Advanced Research Projects Agency; National Science Foundation; Government of Canada; Canadian Institute for Advanced Research; Alfred P. Sloan Foundation","keywords":"Differential privacy; Regularization (linguistics); Focus (optics); Convex optimization; Regular polygon; Optimization problem; Training set; Noise (video)","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.01290132,0.001410158,0.002509165,0.0008918815,0.001771267,0.004591008,0.003660281,0.003332295,0.003543926],"category_scores_gemma":[0.07814011,0.00133555,0.001853189,0.001643585,0.004648935,0.009574047,0.01143538,0.00804676,0.001314118],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.003422688,"about_ca_system_score_gemma":0.005728533,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.001974879,"about_ca_topic_score_gemma":0.002342382,"domain_scores_codex":[0.9808343,0.008838758,0.0008961912,0.003261896,0.004560648,0.00160816],"domain_scores_gemma":[0.9365081,0.03602028,0.003086438,0.0205695,0.002735043,0.001080688],"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":[0.001408716,0.0003557361,0.006176517,0.0002982655,0.0002111711,0.0003861411,0.0006156003,0.5019003,0.01250497,0.3119729,0.01118225,0.1529875],"study_design_scores_gemma":[0.00006049039,0.00004520051,0.0003047134,0.00002139126,0.0000128089,0.0000592155,0.00003626041,0.8350844,0.004961958,0.1584334,0.0009598708,0.00002018797],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.03050971,0.000160458,0.9617954,0.002177862,0.00004729237,0.0001080086,0.000379224,0.002081586,0.002740382],"genre_scores_gemma":[0.7675017,0.0001601561,0.2258496,0.001053736,0.0001532556,0.0003020249,0.001000403,0.0006776553,0.003301572],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.01290132,"threshold_uncertainty_score":0.0682295,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.05848234471022681,"score_gpt":0.2817977020491695,"score_spread":0.2233153573389427,"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."}}