{"id":"W3172136722","doi":"10.18653/v1/2021.acl-long.144","title":"Causal Analysis of Syntactic Agreement Mechanisms in Neural Language Models","year":2021,"lang":"en","type":"preprint","venue":"","topic":"Topic Modeling","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Azrieli Foundation; National Science Foundation","keywords":"Agreement; Sentence; Computer science; Verb; Natural language processing; Artificial intelligence; Linguistics; Subject (documents); Sentence processing; Syntax","routes":{"ca_aff":false,"ca_fund":true,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.004086672,0.0004794441,0.001154567,0.001573471,0.0009294827,0.002281501,0.001952371,0.001398721,0.007100613],"category_scores_gemma":[0.02778346,0.0009878451,0.001219524,0.001080868,0.001481265,0.00558905,0.002060013,0.002447844,0.000543422],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001425263,"about_ca_system_score_gemma":0.0008443169,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.004074625,"about_ca_topic_score_gemma":0.004596691,"domain_scores_codex":[0.9988932,0.0006487779,0.00005400492,0.000175196,0.0001368724,0.0000919775],"domain_scores_gemma":[0.9760439,0.02079959,0.0008758754,0.000951591,0.0009189309,0.0004100309],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0002413028,0.00008423439,0.003311416,0.0001432152,0.0001839604,0.0001844214,0.0004275134,0.1796331,0.001804647,0.7851173,0.003376856,0.02549218],"study_design_scores_gemma":[0.00002117396,0.00001109165,0.0004259919,0.00001146178,0.0000294486,0.00002333694,0.0000318493,0.6026716,0.0002933845,0.3960916,0.0003735823,0.00001556121],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.1983732,0.001752152,0.7807007,0.00529548,0.0001810638,0.00005269794,0.0005354558,0.001048451,0.0120609],"genre_scores_gemma":[0.966223,0.0005692551,0.02920825,0.0001588616,0.000150724,0.00008280494,0.0003258016,0.0002638914,0.003017391],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.007100613,"threshold_uncertainty_score":0.02375394,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03695777655512643,"score_gpt":0.2821153186159529,"score_spread":0.2451575420608265,"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."}}