{"id":"W4406986127","doi":"10.32388/pjil3e","title":"SafeSynthDP: Leveraging Large Language Models for Privacy-Preserving Synthetic Data Generation Using Differential Privacy","year":2025,"lang":"en","type":"preprint","venue":"Qeios","topic":"Privacy-Preserving Technologies in Data","field":"Computer Science","cited_by":4,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Alberta","funders":"","keywords":"Differential privacy; Computer science; Internet privacy; Computer security; Information privacy; Privacy software; Privacy protection; Data mining","routes":{"ca_aff":true,"ca_fund":false,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":["metaresearch","metaepi_narrow","scholarly_communication","open_science"],"consensus_categories":["open_science"],"category_scores_codex":[0.001340811,0.0008143089,0.0008980205,0.0006436845,0.0005877605,0.001368713,0.09907749,0.000778348,0.00004659771],"category_scores_gemma":[0.02160036,0.0008720443,0.0002453581,0.000464628,0.00008360523,0.002205386,0.5572785,0.00125845,0.000008530664],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0004434853,"about_ca_system_score_gemma":0.0006273983,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0002541757,"about_ca_topic_score_gemma":0.00003628009,"domain_scores_codex":[0.9934297,0.0002971694,0.0009926055,0.003255464,0.0008115603,0.001213501],"domain_scores_gemma":[0.9546345,0.0004712638,0.0006048052,0.04391172,0.0002435454,0.0001342013],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"not_applicable","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.000161178,0.001960402,0.0004051285,0.008890642,0.002462268,0.0002651528,0.008754773,0.03748978,0.05365438,0.03022522,0.6913569,0.1643742],"study_design_scores_gemma":[0.0004999306,0.00001411371,0.00001417358,0.0007381123,0.0001185554,0.000009285007,0.00004264681,0.9028441,0.005645495,0.08814414,0.001168083,0.0007613579],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.02394635,0.001048381,0.9604036,0.005750515,0.002737415,0.001835978,0.002344737,0.001672923,0.0002600985],"genre_scores_gemma":[0.3733031,0.0001188487,0.6236719,0.0001740428,0.0005150398,0.0001816984,0.001817185,0.00007042097,0.0001477144],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.8653544,"threshold_uncertainty_score":0.9996679,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1236772230759604,"score_gpt":0.3420301906008161,"score_spread":0.2183529675248558,"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."}}