{"id":"W4389518424","doi":"10.18653/v1/2023.arabicnlp-1.9","title":"Beyond English: Evaluating LLMs for Arabic Grammatical Error Correction","year":2023,"lang":"en","type":"article","venue":"","topic":"Natural Language Processing Techniques","field":"Computer Science","cited_by":14,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of British Columbia","funders":"Alliance de recherche numérique du Canada; Social Sciences and Humanities Research Council of Canada; Natural Sciences and Engineering Research Council of Canada; Canada Research Chairs","keywords":"Computer science; Arabic; Context (archaeology); Natural language processing; Artificial intelligence; Task (project management); Linguistics","routes":{"ca_aff":true,"ca_fund":true,"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":[0.000842099,0.0001030886,0.0001178308,0.0001384113,0.0001488068,0.0001940342,0.0005057966,0.00007453873,0.00001319358],"category_scores_gemma":[0.001421586,0.00008284754,0.0000592114,0.0006657263,0.00002283915,0.0004536879,0.0001711777,0.0001267576,0.0000406472],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00004155434,"about_ca_system_score_gemma":0.00004682684,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.000009991005,"about_ca_topic_score_gemma":0.000007129197,"domain_scores_codex":[0.9989032,0.00003905412,0.0001927918,0.0003134958,0.0002732979,0.0002781497],"domain_scores_gemma":[0.9989998,0.0003461617,0.00005953158,0.0002954194,0.0002452739,0.00005379284],"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":[0.00001357544,0.00004545127,0.00009515298,0.00009217292,0.00001626534,0.000007550698,0.003461581,0.00007333726,0.007288242,0.2365107,0.04860682,0.7037891],"study_design_scores_gemma":[0.0001715501,0.0001524978,0.00006221679,0.00002829838,0.000006508739,0.00000655939,0.00009392632,0.7737695,0.02423837,0.200872,0.0004233072,0.0001752986],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.005621244,0.00007629366,0.9860058,0.0006931844,0.001580971,0.0003567007,7.872194e-7,0.004086529,0.001578442],"genre_scores_gemma":[0.1855459,0.000001130475,0.8115435,0.0002491291,0.0001802234,0.0001271153,0.000009335484,0.00001300716,0.002330689],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.7736961,"threshold_uncertainty_score":0.3378425,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.04045139428244435,"score_gpt":0.3526177338742187,"score_spread":0.3121663395917744,"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."}}