{"id":"W4393779445","doi":"10.5281/zenodo.8381476","title":"MatSciML: A Broad, Multi-Task Benchmark for Solid-State Materials Modeling","year":2023,"lang":"en","type":"dataset","venue":"Zenodo (CERN European Organization for Nuclear Research)","topic":"Machine Learning in Materials Science","field":"Materials Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Vector Institute","funders":"","keywords":"Benchmark (surveying); Task (project management); Solid-state; Computer science; Engineering; Systems engineering; Geography; Cartography; Engineering 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":["metaepi_narrow","sts","scholarly_communication","insufficient_payload"],"consensus_categories":["insufficient_payload"],"category_scores_codex":[0.004456789,0.00056736,0.0007036431,0.0005742543,0.003584798,0.004201353,0.004096351,0.0002813306,0.01691659],"category_scores_gemma":[0.003246417,0.0005807737,0.0001343661,0.0006260464,0.0003316927,0.0004590113,0.003632035,0.0004463647,0.0712761],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0002597971,"about_ca_system_score_gemma":0.00002604658,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0003991569,"about_ca_topic_score_gemma":0.000005311876,"domain_scores_codex":[0.9942787,0.0009762461,0.001007093,0.001522095,0.0009706158,0.001245242],"domain_scores_gemma":[0.9966677,0.0001136191,0.0005877237,0.001420405,0.0008723442,0.0003382085],"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":[0.0001093333,0.00009053701,3.491558e-8,0.0005558691,0.00002217841,0.00002029487,0.0001949872,0.005105323,0.1448022,0.00002736103,0.8485566,0.0005152894],"study_design_scores_gemma":[0.0006747856,0.0002450069,0.000006375468,0.000212677,0.00005054482,0.00007619563,0.00006204629,0.01064309,0.003769997,0.0002041398,0.9833824,0.0006727573],"study_design_candidate":"not_applicable","study_design_consensus":"not_applicable","genre_codex":"dataset","genre_gemma":"dataset","genre_scores_codex":[0.003723499,0.00003540346,0.01111516,0.0002269891,0.001839555,0.001533982,0.9796754,0.001626583,0.0002234631],"genre_scores_gemma":[0.000875516,0.0001710862,0.002954244,0.0001841137,0.00057602,0.000001740508,0.991705,0.002535399,0.0009968708],"genre_candidate":"dataset","genre_consensus":"dataset","teacher_disagreement_score":0.1410322,"threshold_uncertainty_score":0.9996644,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.04824776266788908,"score_gpt":0.3044233479694183,"score_spread":0.2561755853015292,"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."}}