{"id":"W6891789982","doi":"10.48448/1jt2-gg52","title":"Towards Understanding Large-Scale Discourse Structures in Pre-Trained and Fine-Tuned Language Models","year":2022,"lang":"en","type":"other","venue":"Underline Science Inc.","topic":"","field":"","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of British Columbia","funders":"","keywords":"Variety (cybernetics); Discourse analysis; Domain of discourse; Information structure; Language understanding; Language model","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":[0.004728804,0.001279295,0.0007877718,0.001443804,0.000684657,0.003140267,0.001647374,0.001894168,0.002619968],"category_scores_gemma":[0.02108282,0.0008557396,0.0007705688,0.0008188359,0.0009649608,0.004707248,0.001923154,0.004358145,0.001390439],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.002175521,"about_ca_system_score_gemma":0.001572439,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.01012068,"about_ca_topic_score_gemma":0.02149154,"domain_scores_codex":[0.9975707,0.001387988,0.00007656724,0.0006491941,0.0001959802,0.0001195599],"domain_scores_gemma":[0.9855161,0.01124903,0.0005864093,0.001726951,0.0006336899,0.0002878275],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"not_applicable","study_design_scores_codex":[0.0007105296,0.0004169462,0.0102561,0.0003484757,0.0004681392,0.0002484312,0.001694148,0.6767961,0.01905039,0.01725483,0.00665122,0.2661047],"study_design_scores_gemma":[0.00001805447,0.00004741628,0.00110967,0.00003044884,0.00002020235,0.00002107258,0.0001207738,0.9815603,0.002037498,0.01367832,0.001336463,0.00001978377],"study_design_candidate":"not_applicable","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"other","genre_scores_codex":[0.2574619,0.002244643,0.7250955,0.002458093,0.0001810127,0.0001492304,0.001834318,0.004007418,0.006567963],"genre_scores_gemma":[0.8245686,0.0003772774,0.165194,0.0004700874,0.0001263413,0.0001813467,0.003817749,0.0005149244,0.004749553],"genre_candidate":"other","genre_consensus":null,"teacher_disagreement_score":0.01012068,"threshold_uncertainty_score":0.02500862,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03144478449659047,"score_gpt":0.3239771923018189,"score_spread":0.2925324078052284,"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."}}