{"id":"W4403736505","doi":"10.1093/ehjdh/ztae079","title":"Fine-tuned large language models can generate expert-level echocardiography reports","year":2024,"lang":"en","type":"editorial","venue":"European Heart Journal - Digital Health","topic":"Cardiovascular Function and Risk Factors","field":"Medicine","cited_by":3,"is_retracted":false,"has_abstract":false,"ca_institutions":"Université de Montréal; Montreal Heart Institute","funders":"Institut de Cardiologie de Montréal; Fondation Institut de Cardiologie de Montréal; Canadian Institute for Advanced Research","keywords":"Computer science; Natural language processing","routes":{"ca_aff":true,"ca_fund":true,"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","research_integrity"],"consensus_categories":[],"category_scores_codex":[0.002525659,0.0007678359,0.001617399,0.0008736313,0.0004390324,0.0008717826,0.0001871522,0.0003619402,0.00007538571],"category_scores_gemma":[0.0004222389,0.0006015089,0.002843898,0.0006020683,0.00007457766,0.0003338387,0.0001938251,0.003457451,0.0002614108],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0005453179,"about_ca_system_score_gemma":0.002040444,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0001640854,"about_ca_topic_score_gemma":0.00003704884,"domain_scores_codex":[0.9936792,0.0005220947,0.001554999,0.0009814646,0.002170737,0.001091544],"domain_scores_gemma":[0.9962846,0.0001263202,0.0004814516,0.001059026,0.0004668627,0.001581759],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"not_applicable","study_design_gemma":"not_applicable","study_design_scores_codex":[0.0001065747,0.0001248997,0.000176614,0.0003118142,0.0008476183,0.01380964,0.0008484239,0.00003199777,0.000008405183,0.000002327527,0.9775906,0.006141147],"study_design_scores_gemma":[0.001103959,0.0005145749,0.0003262665,0.001276346,0.0001701361,0.006577796,0.00026796,0.00002599718,0.000009824301,0.00002963578,0.989168,0.0005294958],"study_design_candidate":"not_applicable","study_design_consensus":"not_applicable","genre_codex":"editorial","genre_gemma":"editorial","genre_scores_codex":[0.003710325,0.03318209,0.000932849,0.001704619,0.9484254,0.0008041736,0.002222911,0.000562564,0.008455072],"genre_scores_gemma":[0.04051431,0.005332167,0.0002597808,0.001546626,0.9348028,0.000009322873,0.002856393,0.0006118748,0.01406669],"genre_candidate":"editorial","genre_consensus":"editorial","teacher_disagreement_score":0.03680399,"threshold_uncertainty_score":0.9996436,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03605042373573915,"score_gpt":0.3149609740743287,"score_spread":0.2789105503385896,"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."}}