{"id":"W4406486231","doi":"10.1007/978-3-031-88711-6_10","title":"Evaluating LLM Abilities to Understand Tabular Electronic Health Records: A Comprehensive Study of Patient Data Extraction and Retrieval","year":2025,"lang":"en","type":"book-chapter","venue":"Lecture notes in computer science","topic":"Machine Learning in Healthcare","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":false,"ca_institutions":"","funders":"Fonds de Recherche du Québec - Santé; Grand Équipement National De Calcul Intensif","keywords":"Computer science; Task (project management); Context (archaeology); Data extraction; Information retrieval; Health records; Selection (genetic algorithm); Serialization; Data mining; Machine learning; Feature selection; Data science; Artificial intelligence; MEDLINE; Health care; Engineering","routes":{"ca_aff":false,"ca_fund":true,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.002006558,0.0004861697,0.0008193689,0.0009844314,0.0004219439,0.0002826043,0.002693904,0.0001798314,0.000006849838],"category_scores_gemma":[0.0006277455,0.0004799393,0.00004593885,0.0009560199,0.0002723277,0.0004844812,0.003346605,0.00130508,0.000001660502],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0009506273,"about_ca_system_score_gemma":0.002339877,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0007214437,"about_ca_topic_score_gemma":0.001142991,"domain_scores_codex":[0.9940318,0.0003753424,0.001017389,0.002315888,0.001489881,0.0007696355],"domain_scores_gemma":[0.9942752,0.00161224,0.0006692482,0.002708249,0.000531044,0.0002040721],"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":[0.00008320953,0.0001810682,0.0004208173,0.0005574152,0.00004334576,0.00001896928,0.01615641,0.06445502,0.00003732871,0.005399623,0.00003526184,0.9126115],"study_design_scores_gemma":[0.0007169433,0.01030321,0.0008501933,0.001517641,0.00002318473,0.00004681135,0.0001053352,0.947687,0.00006076857,0.03704273,0.0008611157,0.000785115],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.03010155,0.002148322,0.9625491,0.001713092,0.001161062,0.002119104,0.00001731745,0.0001078687,0.00008260679],"genre_scores_gemma":[0.7289001,0.0001668989,0.2687925,0.001789931,0.0001854529,0.00001001345,0.00001955808,0.00004055602,0.00009500352],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.9118264,"threshold_uncertainty_score":0.9997652,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.07135074493848703,"score_gpt":0.3813718943281653,"score_spread":0.3100211493896782,"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."}}