{"id":"W6906487774","doi":"10.17605/osf.io/93s5b","title":"Equivalence Checking Benchmark","year":2021,"lang":"en","type":"article","venue":"OSF Preprints (OSF Preprints)","topic":"Logic, programming, and type systems","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of British Columbia","funders":"","keywords":"Benchmark (surveying); Equivalence (formal languages); Formal equivalence checking; Feature (linguistics)","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","insufficient_payload"],"consensus_categories":["insufficient_payload"],"category_scores_codex":[0.003761306,0.0003017912,0.0003897276,0.0001025955,0.000287518,0.0005523038,0.002571206,0.0002013399,0.09948979],"category_scores_gemma":[0.001697088,0.0003218447,0.0002408377,0.0005817519,0.0001139526,0.0006594484,0.003619528,0.0004202324,0.5044012],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001690001,"about_ca_system_score_gemma":0.0003127301,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0001247914,"about_ca_topic_score_gemma":0.00003268657,"domain_scores_codex":[0.9950458,0.0005720464,0.0005719943,0.002523406,0.0006117239,0.0006749773],"domain_scores_gemma":[0.9938073,0.0003166154,0.0002558401,0.005053077,0.0002905345,0.0002766804],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"not_applicable","study_design_scores_codex":[0.00002411743,0.000718682,0.05297368,0.0003355843,0.0002613681,0.0007242649,0.005696537,0.0005700402,0.01698336,0.6626306,0.01849643,0.2405853],"study_design_scores_gemma":[0.001227145,0.000004334082,0.02634573,0.0001031074,0.00006646616,0.0007154607,0.0003656278,0.009226387,0.08812904,0.1591232,0.7131281,0.001565323],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":null,"genre_codex":"other","genre_gemma":"empirical","genre_scores_codex":[0.007335172,0.00001386433,0.3031313,0.0003766397,0.001037557,0.0003537886,9.117334e-7,0.0003451318,0.6874056],"genre_scores_gemma":[0.8447825,0.00006670589,0.009182661,0.0002553976,0.0001512605,0.0001017286,0.0000073021,0.00002331406,0.1454291],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.8374473,"threshold_uncertainty_score":0.9999233,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02505475403201073,"score_gpt":0.2633263971787922,"score_spread":0.2382716431467815,"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."}}