{"id":"W4416118802","doi":"10.48550/arxiv.2504.05227","title":"A Reality Check of Vision-Language Pre-training in Radiology: Have We Progressed Using Text?","year":2025,"lang":"en","type":"preprint","venue":"ArXiv.org","topic":"Multimodal Machine Learning Applications","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Natural Sciences and Engineering Research Council of Canada; Compute Canada","keywords":"Popularity; Feature (linguistics); Reality check; Check-in; Noise (video); Visualization","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":[0.001006471,0.0003397445,0.0006584989,0.0003595008,0.00009949959,0.00006769149,0.001851968,0.0003941982,0.00001115877],"category_scores_gemma":[0.000430279,0.0003447267,0.0001556284,0.0005169035,0.0001617984,0.0001385731,0.002190716,0.001228632,0.00000750432],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001650313,"about_ca_system_score_gemma":0.0006170973,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.003576689,"about_ca_topic_score_gemma":0.0001439901,"domain_scores_codex":[0.997044,0.0004204323,0.0007463363,0.001105339,0.0002624504,0.00042146],"domain_scores_gemma":[0.9969966,0.0003410789,0.0005995727,0.001854647,0.0001169195,0.00009122185],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.00004029984,0.0004074733,0.824469,0.0009211485,0.0001427939,0.0000441626,0.03565949,0.04244831,0.009578376,0.002478191,0.00006259346,0.0837482],"study_design_scores_gemma":[0.0003442137,0.00003412411,0.5766447,0.0007881128,0.00002126079,0.000007958172,0.000166955,0.4197733,0.0006571823,0.001137357,0.0001187701,0.0003060831],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9242219,0.0005731335,0.07111218,0.001929981,0.0002468502,0.0008997536,0.00002338439,0.0002416861,0.0007511043],"genre_scores_gemma":[0.945667,0.0000306503,0.05383336,0.00008017285,0.00008649696,0.0001502439,0.00003020547,0.00001781668,0.0001040708],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.377325,"threshold_uncertainty_score":0.9999005,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.06354629236181726,"score_gpt":0.3889836129961712,"score_spread":0.325437320634354,"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."}}