{"id":"W4389519573","doi":"10.18653/v1/2023.findings-emnlp.98","title":"Dolphin: A Challenging and Diverse Benchmark for Arabic NLG","year":2023,"lang":"en","type":"article","venue":"","topic":"Topic Modeling","field":"Computer Science","cited_by":7,"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; Benchmark (surveying); Automatic summarization; Natural language processing; Modular design; Artificial intelligence; Set (abstract data type); Modern Standard Arabic; Generalization; Arabic; Linguistics; Programming language","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":[0.0002005127,0.00005529901,0.00006615784,0.00007241131,0.00009822447,0.00005790516,0.0002263384,0.0000232731,0.000008191761],"category_scores_gemma":[0.00002339046,0.00004841572,0.00002448137,0.000136385,0.00000840473,0.0002019259,0.0002312297,0.0000327693,0.00002212605],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.000007809649,"about_ca_system_score_gemma":0.000009038302,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.000009618041,"about_ca_topic_score_gemma":0.000005540931,"domain_scores_codex":[0.9994143,0.000006934753,0.00007719477,0.0002368877,0.00008456341,0.0001800929],"domain_scores_gemma":[0.99962,0.00008688345,0.00001440647,0.0002140972,0.00001791638,0.0000467527],"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":[0.00000401897,0.00001856153,0.0009394941,0.00009963189,0.00002494858,0.00002249408,0.004712444,0.001624245,0.0005189698,0.6224809,0.003734245,0.36582],"study_design_scores_gemma":[0.0002229232,0.00002355675,0.0009631297,0.000009876357,0.000001872231,0.000003038629,0.0001466227,0.9818348,0.00006680008,0.01207845,0.004551986,0.00009696936],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.08324706,0.00006014117,0.9101452,0.002627163,0.0002887116,0.0001484749,6.834829e-7,0.0003368601,0.003145707],"genre_scores_gemma":[0.9496628,0.00003744979,0.04828541,0.0002275103,0.00007808197,0.00002000437,0.000001005646,0.00000439695,0.001683363],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.9802105,"threshold_uncertainty_score":0.1974336,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.0487968856808134,"score_gpt":0.2677313356933869,"score_spread":0.2189344500125735,"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."}}