{"id":"W4403289443","doi":"10.29007/hgfv","title":"ARCH-COMP 2024 Category Report: Falsification","year":2024,"lang":"en","type":"article","venue":"EPiC series in computing","topic":"Adversarial Robustness in Machine Learning","field":"Computer Science","cited_by":7,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"ACT-X; JST-Mirai Program; Exploratory Research for Advanced Technology; Electronic Components and Systems for European Leadership; Defense Advanced Research Projects Agency; Universität Salzburg; Alliance de recherche numérique du Canada; Japan Society for the Promotion of Science; Austrian Science Fund; Division of Civil, Mechanical and Manufacturing Innovation; European Commission; Technische Universität Wien; Core Research for Evolutional Science and Technology; National Science Foundation","keywords":"Benchmark (surveying); Arch; Competition (biology); Computer science; Software engineering; Data science; Artificial intelligence; Engineering; Civil engineering; Geography","routes":{"ca_aff":false,"ca_fund":true,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[{"model":"gemma","categories":[],"domain":null,"study_design":"not_applicable","genre":"other","about_ca_system":false,"about_ca_topic":false,"confidence":"low","status":"direct model label, unvalidated"},{"model":"gpt","categories":[],"domain":null,"study_design":"design_other","genre":"other","about_ca_system":false,"about_ca_topic":false,"confidence":"low","status":"direct model label, unvalidated"}],"prediction":{"model_version":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.04038029,0.003741929,0.00218021,0.00407555,0.00411644,0.01221965,0.007083328,0.007540287,0.08704304],"category_scores_gemma":[0.05123863,0.001085427,0.002862496,0.002099521,0.002874791,0.006864425,0.01057529,0.00728207,0.06926835],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.004851815,"about_ca_system_score_gemma":0.008237901,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.008075395,"about_ca_topic_score_gemma":0.009432658,"domain_scores_codex":[0.9552484,0.01042098,0.001810656,0.004478869,0.0236631,0.004378081],"domain_scores_gemma":[0.9099845,0.01994104,0.002004491,0.01880201,0.04326984,0.005998238],"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":[0.00181851,0.0004961974,0.001276975,0.0007315744,0.0001171276,0.0002339847,0.0001596937,0.006689629,0.004297707,0.02168701,0.9021277,0.0603639],"study_design_scores_gemma":[0.0006815284,0.001219603,0.002038118,0.0004625808,0.0001071885,0.0006451206,0.000466735,0.031216,0.02843452,0.03578991,0.8987275,0.0002112594],"study_design_candidate":"not_applicable","study_design_consensus":"not_applicable","genre_codex":"other","genre_gemma":"empirical","genre_scores_codex":[0.04772394,0.006665145,0.317294,0.02792571,0.04418803,0.004682997,0.1060817,0.07858473,0.3668536],"genre_scores_gemma":[0.2099287,0.001955948,0.1610252,0.01022056,0.004570137,0.003521348,0.3832259,0.02982818,0.195724],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.08704304,"threshold_uncertainty_score":0.291188,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01676143807099343,"score_gpt":0.295545062689395,"score_spread":0.2787836246184015,"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."}}