{"id":"W6947969560","doi":"10.48448/s5xp-mn57","title":"Learning to Perturb Word Embeddings for Out-of-distribution QA","year":2021,"lang":"en","type":"other","venue":"Underline Science Inc.","topic":"Memory, History, Trauma, Identity","field":"Arts and Humanities","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Kootenay Association for Science & Technology","funders":"","keywords":"Overfitting; Word embedding; Word (group theory); Training set; Embedding; Context (archaeology); Benchmark (surveying)","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":[],"category_scores_codex":[0.0006599043,0.0002521085,0.000382471,0.000403695,0.0006298511,0.0002779202,0.0005835749,0.0001103788,0.006562275],"category_scores_gemma":[0.0005751604,0.000249685,0.0001261896,0.0001604839,0.001013931,0.0001388698,0.0001275091,0.000251712,0.0001150795],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0002912422,"about_ca_system_score_gemma":0.0003814026,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0006118763,"about_ca_topic_score_gemma":0.007448843,"domain_scores_codex":[0.9979952,0.00003455855,0.0003350647,0.0006181513,0.0005856984,0.0004313166],"domain_scores_gemma":[0.9987036,0.00007102105,0.0003031763,0.0003670632,0.0004036294,0.0001515014],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"not_applicable","study_design_gemma":"not_applicable","study_design_scores_codex":[0.00001906135,0.0001566265,0.00005089939,0.0005301199,0.0000534418,0.000006207592,0.01426669,0.0000515169,0.0009171578,0.07787129,0.8906406,0.01543635],"study_design_scores_gemma":[0.0001677811,0.00010898,0.00001451395,0.0002492284,0.00003841316,8.133531e-7,0.002188667,0.0004278652,0.00008833325,0.0002737152,0.9961376,0.0003041437],"study_design_candidate":"not_applicable","study_design_consensus":"not_applicable","genre_codex":"other","genre_gemma":"other","genre_scores_codex":[0.001950474,0.001006614,0.006944919,0.0005925295,0.01073524,0.001141454,0.0004600626,0.0003834182,0.9767853],"genre_scores_gemma":[0.02408367,0.00002105724,0.002098885,0.0001833888,0.001645144,0.0000391239,0.0003864412,0.0001533894,0.9713889],"genre_candidate":"other","genre_consensus":"other","teacher_disagreement_score":0.1054969,"threshold_uncertainty_score":0.9999955,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.0500092685920365,"score_gpt":0.2971528522932509,"score_spread":0.2471435837012144,"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."}}