{"id":"W4414028042","doi":"10.1101/2025.09.01.673319","title":"BioML-bench: Evaluation of AI Agents for End-to-End Biomedical ML","year":2025,"lang":"en","type":"preprint","venue":"bioRxiv (Cold Spring Harbor Laboratory)","topic":"Scientific Computing and Data Management","field":"Decision Sciences","cited_by":5,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Toronto","funders":"","keywords":"End-to-end principle; Bench to bedside; Computer science; Artificial intelligence; Medicine; Medical physics","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":[0.02380058,0.0004918527,0.0008309216,0.002082574,0.0002668721,0.000779492,0.003071154,0.0003958673,0.0005754504],"category_scores_gemma":[0.01172584,0.0004490228,0.0003422751,0.003089211,0.0002410541,0.0002046912,0.002860615,0.0003966587,0.0001841804],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0003376639,"about_ca_system_score_gemma":0.001913839,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00009195041,"about_ca_topic_score_gemma":0.000007026585,"domain_scores_codex":[0.9897652,0.0005482533,0.001720372,0.00251467,0.00479898,0.0006524714],"domain_scores_gemma":[0.9902511,0.0008678835,0.0008507758,0.004000417,0.003610221,0.0004195604],"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":[0.0003077286,0.002558121,0.009998582,0.001595319,0.001538384,0.0000360737,0.000313737,0.009226668,0.1417508,0.01813976,0.8062344,0.008300434],"study_design_scores_gemma":[0.002935529,0.0002650153,0.09728544,0.001396637,0.001155956,1.43007e-8,0.00005263206,0.1923223,0.06604628,0.0008004975,0.6358144,0.001925196],"study_design_candidate":"not_applicable","study_design_consensus":"not_applicable","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.6706843,0.0007674043,0.2820562,0.005506149,0.02686381,0.006371624,0.006994017,0.0004994634,0.0002570648],"genre_scores_gemma":[0.9855165,0.00001012876,0.01299748,0.0006144688,0.0003755735,0.0003147711,0.000005135016,0.00003460701,0.0001313638],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.3148322,"threshold_uncertainty_score":0.9997962,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1302289515264287,"score_gpt":0.3852932150268801,"score_spread":0.2550642635004513,"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."}}