{"id":"W4393607455","doi":"10.5281/zenodo.7304210","title":"A Survey on Automated Driving System Testing: Landscapes and Trends [Supplementary material]","year":2022,"lang":"en","type":"dataset","venue":"Zenodo (CERN European Organization for Nuclear Research)","topic":"Autonomous Vehicle Technology and Safety","field":"Engineering","cited_by":1,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Alberta","funders":"","keywords":"Computer science","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":["metaepi_narrow","sts","insufficient_payload"],"consensus_categories":[],"category_scores_codex":[0.0007113077,0.0002618649,0.0002810833,0.0004950707,0.001506525,0.0003172073,0.0008197835,0.000172146,0.04662417],"category_scores_gemma":[0.0001539979,0.0002923357,0.00002825831,0.0005381089,0.00006462412,0.00006073343,0.001141128,0.0005496503,0.0004695999],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0002724327,"about_ca_system_score_gemma":0.000002427725,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00009130956,"about_ca_topic_score_gemma":0.000008227295,"domain_scores_codex":[0.9983194,0.0003494899,0.0003142503,0.0004061558,0.000241388,0.0003692999],"domain_scores_gemma":[0.9991795,0.00004884699,0.00009503841,0.0004961547,0.00008091892,0.00009947881],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"not_applicable","study_design_gemma":"not_applicable","study_design_scores_codex":[0.00001618974,0.00003116676,0.000003794647,0.0001851083,0.00008095071,0.00004160274,0.00003621267,0.0003006503,0.00008545611,0.00002258951,0.9887597,0.01043661],"study_design_scores_gemma":[0.0002594874,0.0002076769,0.00222318,0.00005361754,0.00002945362,0.0001090757,0.00005587839,0.003674991,0.00002714847,0.000001355117,0.9930793,0.0002788134],"study_design_candidate":"not_applicable","study_design_consensus":"not_applicable","genre_codex":"dataset","genre_gemma":"dataset","genre_scores_codex":[0.00436755,0.00003514476,0.00001061266,0.00004098058,0.0002179865,0.0002099725,0.9882496,0.004904296,0.001963875],"genre_scores_gemma":[0.0471909,0.00006814864,0.00002480522,0.00001687286,0.00006028772,1.728722e-7,0.9516601,0.0009342834,0.00004440847],"genre_candidate":"dataset","genre_consensus":"dataset","teacher_disagreement_score":0.04615457,"threshold_uncertainty_score":0.9999529,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02200504708889908,"score_gpt":0.2246830406495441,"score_spread":0.202677993560645,"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."}}