{"id":"W3155600250","doi":"10.18653/v1/2021.naacl-main.138","title":"Modeling Event Plausibility with Consistent Conceptual Abstraction","year":2021,"lang":"en","type":"preprint","venue":"","topic":"Topic Modeling","field":"Computer Science","cited_by":1,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Toronto; McGill University","funders":"Canadian Institute for Advanced Research; Compute Canada; Microsoft Research","keywords":"Computer science; Natural language processing; Transformer; Hierarchy; Artificial intelligence; Cognitive psychology; Psychology","routes":{"ca_aff":true,"ca_fund":true,"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":[],"consensus_categories":[],"category_scores_codex":[0.0003667615,0.0002720068,0.0003434639,0.00005583826,0.0000844715,0.0003385484,0.0006691425,0.0002216727,0.00006862164],"category_scores_gemma":[0.00002870612,0.0002324213,0.0001348331,0.00007791607,0.00004728432,0.0002733557,0.001293807,0.0006876565,0.00001016824],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0002272093,"about_ca_system_score_gemma":0.0006477609,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.001123286,"about_ca_topic_score_gemma":0.0005627405,"domain_scores_codex":[0.9974675,0.00008010203,0.0004814152,0.001163181,0.0005297898,0.0002779956],"domain_scores_gemma":[0.9978884,0.00004669125,0.000137777,0.001506055,0.0003001674,0.0001208931],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.000005321721,0.000064931,0.0001147415,0.00004981488,0.00004910575,0.000031599,0.0009052085,0.9878452,0.00004790875,0.00596496,0.00001029908,0.004910856],"study_design_scores_gemma":[0.0001850113,0.00002157615,0.0001035087,0.0001289418,0.0000165859,0.00002906312,0.0005763854,0.997066,0.0002590191,0.00129601,0.00002026051,0.0002977056],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.3418578,0.0001138908,0.6553972,0.0002997456,0.0005989209,0.000209922,0.000001399724,0.0001922231,0.001328829],"genre_scores_gemma":[0.8282741,0.00001204612,0.1712754,0.0001690669,0.00008034569,0.00003013138,0.00001389446,0.000009766733,0.0001352675],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.4864163,"threshold_uncertainty_score":0.9477864,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.06236140920146336,"score_gpt":0.2777404454129202,"score_spread":0.2153790362114568,"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."}}