{"id":"W4386566490","doi":"10.18653/v1/2023.findings-eacl.74","title":"Improving Prediction Backward-Compatiblility in NLP Model Upgrade with Gated Fusion","year":2023,"lang":"en","type":"article","venue":"","topic":"Domain Adaptation and Few-Shot Learning","field":"Computer Science","cited_by":1,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Waterloo","funders":"","keywords":"Upgrade; Computer science; Regression; Artificial intelligence; Regression analysis; Machine learning; Ensemble forecasting; Baseline (sea); Data mining; Statistics; Mathematics","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":[],"consensus_categories":[],"category_scores_codex":[0.0004714589,0.0001065663,0.0001128883,0.000308917,0.0001094064,0.0001188629,0.0002820837,0.00004944938,0.00002135941],"category_scores_gemma":[0.00003758458,0.00008760226,0.00002525775,0.001527038,0.00002295137,0.000581237,0.0001312453,0.0001716989,0.00008631249],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00005619909,"about_ca_system_score_gemma":0.00007770898,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0001598853,"about_ca_topic_score_gemma":0.0001012328,"domain_scores_codex":[0.9987776,0.00006495779,0.0002148203,0.0003714556,0.0003060286,0.0002651849],"domain_scores_gemma":[0.9994569,0.00004849759,0.00006138975,0.0002997169,0.00005294312,0.00008057887],"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":[0.00008416965,0.0001777616,0.03304323,0.00006900761,0.00001641171,0.00004109819,0.006435234,0.7949132,0.01262251,0.02374916,0.001778856,0.1270694],"study_design_scores_gemma":[0.0004626022,0.00005001693,0.03705855,0.00001547175,0.000001283517,0.000002435578,0.0001449209,0.9611412,0.0003160796,0.0005316286,0.0001742096,0.0001015869],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.2281386,0.000004483446,0.7668292,0.0003877362,0.00008253312,0.0001480207,0.000001072035,0.0007296909,0.003678712],"genre_scores_gemma":[0.9552718,0.000005465208,0.04348713,0.0001533766,0.00001390828,0.00001003321,0.0000125418,0.000008577024,0.001037165],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.7271332,"threshold_uncertainty_score":0.3572317,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02558255835228114,"score_gpt":0.2397477656793226,"score_spread":0.2141652073270414,"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."}}