{"id":"W4386044034","doi":"10.48550/arxiv.2308.09552","title":"Attesting Distributional Properties of Training Data for Machine Learning","year":2023,"lang":"en","type":"preprint","venue":"arXiv (Cornell University)","topic":"Privacy-Preserving Technologies in Data","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Deutsche Forschungsgemeinschaft; European Commission; Government of Ontario","keywords":"Property (philosophy); Computer science; Trustworthiness; Trainer; Inference; Training set; Diversity (politics); Training (meteorology); Population; Machine learning; Artificial intelligence; Computer security; Medicine","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":[0.03933689,0.001024474,0.00190891,0.002256444,0.003497772,0.007623648,0.005071811,0.004755044,0.003575466],"category_scores_gemma":[0.177065,0.00149268,0.002739751,0.00327946,0.00980878,0.0251616,0.01755978,0.01359797,0.002067826],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.002400683,"about_ca_system_score_gemma":0.004331842,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0005342802,"about_ca_topic_score_gemma":0.0003946672,"domain_scores_codex":[0.9404277,0.02866351,0.004603749,0.008289325,0.01558881,0.002426979],"domain_scores_gemma":[0.6785055,0.130347,0.01412098,0.1671881,0.008206221,0.001632233],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"theoretical_or_conceptual","study_design_scores_codex":[0.0006737435,0.0003492358,0.006438434,0.0003329664,0.0002342726,0.0005996787,0.001666927,0.0221514,0.01552029,0.8041098,0.006051433,0.1418719],"study_design_scores_gemma":[0.0000654473,0.0001510611,0.0007245133,0.0001193021,0.0000740687,0.0005998703,0.0002352287,0.09405751,0.02934428,0.8653426,0.009221348,0.00006497031],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":"theoretical_or_conceptual","genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.01747468,0.0002452426,0.9743939,0.003362564,0.0001161499,0.0001483762,0.0002840775,0.001050027,0.002924863],"genre_scores_gemma":[0.7137441,0.0005153518,0.2784009,0.002074686,0.0004992385,0.0006959011,0.0008771284,0.0004885063,0.00270417],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.03933689,"threshold_uncertainty_score":0.2080358,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.3940355269946956,"score_gpt":0.2476563301850472,"score_spread":0.1463791968096484,"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."}}