{"id":"W4384341701","doi":"10.31234/osf.io/4c5et","title":"Comparing the Oxford Digital Multiple Errands Test (OxMET) to a real-life version: convergence, feasibility, and acceptability.","year":2023,"lang":"en","type":"preprint","venue":"","topic":"Stroke Rehabilitation and Recovery","field":"Medicine","cited_by":1,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Stroke Association; Department of Health and Social Care; National Institute for Health and Care Research","keywords":"Test (biology); Chronic stroke; Stroke (engine); Depression (economics); Montreal Cognitive Assessment; Psychology; Physical therapy; Activities of daily living; Gerontology; Cognitive disabilities; Cognitive impairment; Cognition; Medicine; Rehabilitation; Psychiatry","routes":{"ca_aff":false,"ca_fund":false,"ca_venue":false,"about_ca":true,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[{"model":"gemma","categories":[],"domain":null,"study_design":"observational","genre":"empirical","about_ca_system":false,"about_ca_topic":false,"confidence":"low","status":"direct model label, unvalidated"},{"model":"gpt","categories":[],"domain":null,"study_design":"observational","genre":"empirical","about_ca_system":false,"about_ca_topic":false,"confidence":"high","status":"direct model label, unvalidated"}],"prediction":{"model_version":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.004814649,0.0004799707,0.0004925095,0.001379931,0.0002027282,0.0009004698,0.0005238134,0.0005396781,0.001619534],"category_scores_gemma":[0.02026431,0.0002281238,0.0005931536,0.0007123018,0.000454581,0.001255637,0.001160447,0.0003860325,0.0003875036],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0002757356,"about_ca_system_score_gemma":0.0003342015,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002764948,"about_ca_topic_score_gemma":0.004042664,"domain_scores_codex":[0.9966608,0.001050609,0.0005382858,0.0004563128,0.001134607,0.0001592632],"domain_scores_gemma":[0.9911774,0.003335164,0.002212672,0.0005330609,0.002304406,0.0004373227],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.0005804895,0.0001987424,0.9802556,0.00005478316,0.0001759635,0.00007548923,0.0005936635,0.0002425952,0.0004269193,0.00008954895,0.0004605966,0.01684551],"study_design_scores_gemma":[0.00004790774,0.0006208604,0.9958871,0.00003662437,0.00004130788,0.0003503162,0.0005798604,0.001033972,0.0004349212,0.0001572839,0.0007920735,0.00001773889],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9967492,0.0001645864,0.00108969,0.00003379161,0.00003388632,0.00007025563,0.0002355102,0.00002019395,0.001602944],"genre_scores_gemma":[0.9974042,0.00007138465,0.001247363,0.00003548906,0.00001357505,0.00009037455,0.0005572928,0.00001257656,0.0005677744],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.004814649,"threshold_uncertainty_score":0.02546263,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.08454001528372139,"score_gpt":0.3284089627219942,"score_spread":0.2438689474382728,"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."}}