{"id":"W6939217260","doi":"10.60692/e3qnm-hc398","title":"On Orthogonality Constraints for Transformers","year":2021,"lang":"en","type":"article","venue":"Greater South Information System","topic":"Natural Language Processing Techniques","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Université de Montréal","funders":"","keywords":"Orthogonality; Transformer; Joint (building); Natural language; Algebra over a field; Matrix algebra","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":[0.00528673,0.001087292,0.001180015,0.002140897,0.002594284,0.004036316,0.001920197,0.001576332,0.02581514],"category_scores_gemma":[0.01771799,0.001375393,0.002003368,0.003059211,0.003319727,0.02217624,0.006516841,0.004674249,0.003682186],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.002019038,"about_ca_system_score_gemma":0.002167583,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.004145188,"about_ca_topic_score_gemma":0.007013585,"domain_scores_codex":[0.994779,0.001651528,0.0005229505,0.001001436,0.001290546,0.0007544822],"domain_scores_gemma":[0.9901413,0.005960724,0.0004030868,0.001468594,0.00168567,0.0003406814],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"theoretical_or_conceptual","study_design_scores_codex":[0.00004926549,0.00002638325,0.0001763574,0.00006680888,0.00001433232,0.00007966511,0.0001174844,0.001394864,0.0003840087,0.9723241,0.008699218,0.01666752],"study_design_scores_gemma":[0.00002175468,0.0000103304,0.00006019128,0.00003482006,0.00001179223,0.00007318927,0.00007243724,0.004689405,0.000496025,0.9797881,0.01472923,0.00001268759],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":"theoretical_or_conceptual","genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.03080557,0.002243678,0.8176918,0.006846021,0.001233166,0.000318812,0.002221436,0.001359562,0.13728],"genre_scores_gemma":[0.5501829,0.003873871,0.3888822,0.003832871,0.001804643,0.0005358,0.005643479,0.001939998,0.04330429],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.02581514,"threshold_uncertainty_score":0.08636022,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02878086650117915,"score_gpt":0.245701246738939,"score_spread":0.2169203802377599,"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."}}