{"id":"W6929418907","doi":"10.48448/g3qz-q748","title":"Multilingual Nonce Dependency Treebanks: Understanding how Language Models Represent and Process Syntactic Structure","year":2024,"lang":"en","type":"other","venue":"Underline Science Inc.","topic":"Freshwater macroinvertebrate diversity and ecology","field":"Environmental Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Dalhousie University","funders":"","keywords":"Cryptographic nonce; Perplexity; Grammaticality; Syntax; Dependency (UML); Word (group theory); Argument (complex analysis); Lexicalization","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":[0.004083345,0.0009197096,0.0006180363,0.002120178,0.0008298987,0.00341936,0.001646471,0.001003664,0.005534711],"category_scores_gemma":[0.01522074,0.001186574,0.001130844,0.002170962,0.001245851,0.01251536,0.002749074,0.00233298,0.001819003],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001498349,"about_ca_system_score_gemma":0.001586908,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00872413,"about_ca_topic_score_gemma":0.01220223,"domain_scores_codex":[0.9985791,0.0007196271,0.0001253744,0.0003341978,0.000184994,0.00005671631],"domain_scores_gemma":[0.993681,0.004038635,0.0004458484,0.001206775,0.0005185307,0.0001092511],"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":[0.0003028888,0.0001740876,0.01605087,0.000828895,0.0002882717,0.0008140683,0.00794246,0.06876328,0.01629303,0.5144333,0.02770077,0.3464081],"study_design_scores_gemma":[0.00003019958,0.00004039355,0.004695983,0.0002733716,0.00006378841,0.0003069576,0.001016605,0.4310819,0.008234168,0.4833171,0.07083212,0.000107376],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.02128836,0.0003936796,0.9636639,0.001273036,0.00007018355,0.00009061913,0.005331689,0.003299716,0.004588821],"genre_scores_gemma":[0.362244,0.0009910159,0.6172364,0.00047844,0.00006716618,0.0003516257,0.01439447,0.001769631,0.002467247],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.00872413,"threshold_uncertainty_score":0.02159506,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02489636936238457,"score_gpt":0.2618803352903604,"score_spread":0.2369839659279758,"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."}}