{"id":"W6947849807","doi":"10.48448/zkhk-xm79","title":"EvoGrad: An Online Platform for an Evolving Winograd Schema Challenge using Adversarial Human Perturbations","year":2022,"lang":"en","type":"other","venue":"Underline Science Inc.","topic":"Evolution and Paleontology Studies","field":"Earth and Planetary Sciences","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"McGill University; Brock University","funders":"","keywords":"Adversarial system; Schema (genetic algorithms); Task (project management); Metric (unit); Stability (learning theory); Measure (data warehouse)","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.004189724,0.001720972,0.000771556,0.0007268937,0.0006118434,0.001608469,0.004186051,0.002668968,0.01192428],"category_scores_gemma":[0.01367109,0.0006442972,0.0009641906,0.0004266953,0.001457383,0.003644004,0.006056107,0.003727863,0.004074654],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001063612,"about_ca_system_score_gemma":0.001491852,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.004553656,"about_ca_topic_score_gemma":0.008958544,"domain_scores_codex":[0.9980158,0.00098588,0.00008272861,0.0004556892,0.0003358766,0.000124141],"domain_scores_gemma":[0.996276,0.002146636,0.0001587264,0.0008878871,0.0002916231,0.0002392464],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"not_applicable","study_design_scores_codex":[0.001455581,0.0008708318,0.004962217,0.0008611756,0.0003755713,0.0005601752,0.0007659308,0.6256079,0.006879947,0.04401632,0.1056487,0.2079957],"study_design_scores_gemma":[0.00009958165,0.000120011,0.0001884065,0.00003324833,0.00001097095,0.00006610417,0.00006827842,0.9581975,0.003251543,0.02384384,0.01409308,0.00002736102],"study_design_candidate":"not_applicable","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"other","genre_scores_codex":[0.1090439,0.00201225,0.7696635,0.004170468,0.001437127,0.0009434155,0.006245479,0.07986929,0.02661466],"genre_scores_gemma":[0.528356,0.0004872553,0.4389863,0.00155215,0.0001419275,0.001088292,0.01123879,0.004535139,0.01361425],"genre_candidate":"other","genre_consensus":null,"teacher_disagreement_score":0.01192428,"threshold_uncertainty_score":0.03989065,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1498071195975867,"score_gpt":0.3406532131337748,"score_spread":0.1908460935361881,"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."}}