{"id":"W4409741900","doi":"10.5220/0013491000003932","title":"Banking Strategies and Software Solutions for Generated Test Items","year":2025,"lang":"en","type":"article","venue":"","topic":"Software System Performance and Reliability","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":false,"ca_institutions":"University of Alberta","funders":"","keywords":"Computer science; Test (biology); Software; Software engineering; Operating system","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":[],"consensus_categories":[],"category_scores_codex":[0.0002831133,0.00008793038,0.0001127288,0.00007347766,0.0003794083,0.0003394469,0.0002451232,0.00006106061,0.000004855639],"category_scores_gemma":[0.0001126713,0.00006755966,0.00003412509,0.0003082434,0.00004283461,0.0006025344,0.0001463079,0.00004960834,0.000004331287],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00002373948,"about_ca_system_score_gemma":0.0001871213,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00005652845,"about_ca_topic_score_gemma":0.00004165765,"domain_scores_codex":[0.9992391,0.00001420021,0.0001718462,0.0002642567,0.00006910598,0.0002415085],"domain_scores_gemma":[0.999208,0.0003083774,0.00002826406,0.0002913826,0.0001335517,0.00003044294],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.00001356175,0.0002861753,0.1103636,0.0009861003,0.0001216643,0.000003389286,0.001376669,0.002301307,0.003594563,0.643705,0.03057535,0.2066727],"study_design_scores_gemma":[0.002419547,0.0004300735,0.09084403,0.0003813022,0.00005384086,0.00003369488,0.001233683,0.7294796,0.005407006,0.1240701,0.04452858,0.001118513],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.01958942,0.0004672622,0.9772099,0.0005981711,0.0004157412,0.0002599856,0.000003185791,0.0004008471,0.00105551],"genre_scores_gemma":[0.9402147,0.00002035667,0.05851414,0.00025047,0.00004667831,0.00007293994,0.000003838645,0.000003629224,0.000873249],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.9206253,"threshold_uncertainty_score":0.3273295,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01966233747416383,"score_gpt":0.2610456326138477,"score_spread":0.2413832951396838,"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."}}