{"id":"W3087153552","doi":"10.1177/0265532220957298","title":"Hanyu Shuiping Kaoshi (HSK): A multi-level, multi-purpose proficiency test","year":2020,"lang":"en","type":"article","venue":"Language Testing","topic":"Psychometric Methodologies and Testing","field":"Decision Sciences","cited_by":50,"is_retracted":false,"has_abstract":true,"ca_institutions":"Queen's University","funders":"","keywords":"Language proficiency; Test (biology); Psychology; Argument (complex analysis); Language assessment; Scale (ratio); Mathematics education; Linguistics; Geography","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":["metaresearch","metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.005035285,0.0003886459,0.0006416252,0.0004224829,0.0005002351,0.0005397485,0.001728567,0.000150213,0.0003227926],"category_scores_gemma":[0.5866742,0.0002983867,0.0001700977,0.006194631,0.0001416515,0.0004147852,0.0006769855,0.0005704049,0.0004743075],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00005077856,"about_ca_system_score_gemma":0.0001198002,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0001810563,"about_ca_topic_score_gemma":0.00002195938,"domain_scores_codex":[0.9948754,0.0004547962,0.001214768,0.001278155,0.001273394,0.0009034931],"domain_scores_gemma":[0.9553905,0.04188103,0.0008040175,0.0009346224,0.0004909616,0.0004988471],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"observational","study_design_scores_codex":[0.00001202431,0.000205641,0.3600578,0.0000548075,0.00001276255,0.0003469287,0.007743378,0.000233849,0.05963743,0.0000365781,0.0007123502,0.5709465],"study_design_scores_gemma":[0.003430956,0.0007012363,0.4814237,0.0003851758,0.00006025032,0.0002136085,0.02398005,0.4791605,0.006773615,0.000457817,0.001710408,0.001702705],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.6171786,0.003029125,0.3663179,0.001247611,0.0008284577,0.0009753024,0.00008678201,0.001484194,0.008851971],"genre_scores_gemma":[0.5595,0.000002265452,0.4384884,0.001054371,0.000295155,0.00002014056,0.000003002021,0.00003422099,0.0006024119],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.5816389,"threshold_uncertainty_score":0.9999468,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.8227491867455936,"score_gpt":0.5099737195301289,"score_spread":0.3127754672154647,"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."}}