{"id":"W7082293784","doi":"10.48448/4a71-5066","title":"LLMs can Perform Multi-Dimensional Analytic Writing Assessments: A Case Study of L2 Graduate-Level Academic English Writing","year":2025,"lang":"en","type":"other","venue":"Underline Science Inc.","topic":"Geochemistry and Geologic Mapping","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Saskatchewan","funders":"","keywords":"Context (archaeology); Quality (philosophy); Task (project management); Code (set theory); Academic writing; Graduate students","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":"codex-gemma-dda1882f352a","candidate_categories":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.002428554,0.0004892481,0.0006375682,0.0008946406,0.0005755157,0.0002002512,0.002427761,0.0003002336,0.0000933917],"category_scores_gemma":[0.00093216,0.0004612758,0.00009499857,0.002141378,0.0005169081,0.0003573228,0.001727631,0.00116168,0.000005951742],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001950259,"about_ca_system_score_gemma":0.001540898,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002548277,"about_ca_topic_score_gemma":0.0006704126,"domain_scores_codex":[0.9953464,0.0001115359,0.0009291149,0.001474546,0.001294008,0.0008444146],"domain_scores_gemma":[0.996683,0.0002910513,0.0007597556,0.001069419,0.0009612377,0.0002355596],"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":[0.00008430435,0.01816622,0.1415868,0.00792198,0.003169648,0.02824003,0.06071881,0.03737347,0.00973554,0.04599487,0.08609008,0.5609182],"study_design_scores_gemma":[0.002823667,0.0005181581,0.001117493,0.001910536,0.0001938254,0.0009050965,0.03874829,0.9492588,0.000728295,0.0005461989,0.001490545,0.001759116],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.5141984,0.0009844736,0.1080039,0.002317547,0.005952723,0.005613122,0.0004066558,0.002806139,0.3597171],"genre_scores_gemma":[0.9342651,0.00001189465,0.03261282,0.0001703737,0.0002273005,0.00002776276,0.00001672593,0.00002940601,0.03263865],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.9118853,"threshold_uncertainty_score":0.9997839,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1016390655937194,"score_gpt":0.3582165939705774,"score_spread":0.256577528376858,"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."}}