{"id":"W6891699754","doi":"10.4230/lipics.sat.2024.27","title":"Revisiting SATZilla Features in 2024","year":2024,"lang":"en","type":"article","venue":"DROPS (Schloss Dagstuhl – Leibniz Center for Informatics)","topic":"Formal Methods in Verification","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of British Columbia","funders":"","keywords":"Satisfiability; Preprocessor; Feature selection; Benchmark (surveying); Feature (linguistics); Boolean satisfiability problem; Ranging; Software","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","scholarly_communication"],"consensus_categories":[],"category_scores_codex":[0.001745746,0.0003570867,0.0003761172,0.0005821603,0.0001851855,0.001153882,0.001362346,0.0002369022,0.00002771071],"category_scores_gemma":[0.0002885216,0.0003205797,0.0002195343,0.001010365,0.00007791682,0.002925576,0.0004676607,0.0006909779,0.0002851632],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0002405975,"about_ca_system_score_gemma":0.0001067386,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00001553146,"about_ca_topic_score_gemma":0.000006845614,"domain_scores_codex":[0.9971529,0.00007821296,0.001121108,0.0004007198,0.0005060698,0.0007409505],"domain_scores_gemma":[0.9983585,0.0002446278,0.0002070103,0.0009056254,0.0001386791,0.0001455429],"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":[0.00004473307,0.00009565643,0.001633889,0.002240009,0.0001097623,0.00003764206,0.01686833,0.0003315362,0.0002382529,0.6329607,0.0122637,0.3331758],"study_design_scores_gemma":[0.001343747,0.0001976355,0.003827115,0.001580502,0.00003295739,0.000307872,0.001118868,0.5556071,0.003251901,0.007877544,0.4237337,0.001121005],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.02962401,0.00138947,0.9203405,0.001440623,0.006207416,0.001705502,0.0001844646,0.001007147,0.03810082],"genre_scores_gemma":[0.281825,0.000190048,0.713791,0.001308369,0.0005982094,0.0002409638,0.0001443988,0.00007752216,0.001824447],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.6250832,"threshold_uncertainty_score":0.9999246,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01815246545251966,"score_gpt":0.3067046303418024,"score_spread":0.2885521648892828,"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."}}