{"id":"W6891640368","doi":"10.48448/455f-ep51","title":"How is BERT surprised? Layerwise detection of linguistic anomalies","year":2021,"lang":"en","type":"other","venue":"Underline Science Inc.","topic":"","field":"","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Toronto","funders":"","keywords":"Grammaticality; Language model; Security token; Transformer; Judgement; Gaussian; Word (group theory)","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":[0.001584331,0.0008620544,0.0005349172,0.000827406,0.0004426628,0.001977421,0.0009325836,0.00102458,0.001870818],"category_scores_gemma":[0.009208869,0.0003303335,0.000531648,0.0005331943,0.0006821033,0.002902079,0.001230086,0.001449545,0.001411876],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0008375887,"about_ca_system_score_gemma":0.001078084,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.01242634,"about_ca_topic_score_gemma":0.01417709,"domain_scores_codex":[0.9991146,0.0002390537,0.0000412308,0.0002928121,0.0001810845,0.0001313173],"domain_scores_gemma":[0.9976074,0.0009191812,0.0003078114,0.0004862323,0.000488193,0.0001912189],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.001902266,0.000222302,0.2884601,0.000262349,0.0005654662,0.001592404,0.001584178,0.1446716,0.06573685,0.01580105,0.02385083,0.4553505],"study_design_scores_gemma":[0.00001537513,0.00009302825,0.02524278,0.0000299015,0.0000827663,0.0003575661,0.0004240054,0.9286978,0.0180271,0.02332167,0.00365091,0.0000569837],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.7928303,0.0009053839,0.1829289,0.003882036,0.00024121,0.00003707014,0.002113309,0.00956619,0.007495657],"genre_scores_gemma":[0.9793427,0.0001075082,0.01731191,0.0002223942,0.0000292896,0.000009009413,0.001141119,0.0002217286,0.001614411],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.01242634,"threshold_uncertainty_score":0.02470803,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02351835849607068,"score_gpt":0.2815819742120373,"score_spread":0.2580636157159666,"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."}}