{"id":"W4376653681","doi":"10.48550/arxiv.2305.08264","title":"MatSci-NLP: Evaluating Scientific Language Models on Materials Science Language Tasks Using Text-to-Schema Modeling","year":2023,"lang":"en","type":"preprint","venue":"arXiv (Cornell University)","topic":"Topic Modeling","field":"Computer Science","cited_by":4,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Natural Sciences and Engineering Research Council of Canada; Canadian Institute for Advanced Research","keywords":"Computer science; Artificial intelligence; Natural language processing; Schema (genetic algorithms); Benchmark (surveying); Question answering; Language model; Natural language understanding; Information retrieval; Natural language","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","scholarly_communication"],"consensus_categories":[],"category_scores_codex":[0.003376883,0.0005619006,0.0005744238,0.001534949,0.0009426328,0.001540681,0.004712559,0.0002825662,0.00002785867],"category_scores_gemma":[0.0002609376,0.0006484701,0.0001853357,0.002399282,0.00024119,0.00115443,0.006696954,0.0006254495,0.0002838582],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0007770626,"about_ca_system_score_gemma":0.0009321341,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0008968273,"about_ca_topic_score_gemma":0.00003861752,"domain_scores_codex":[0.9941797,0.0002145325,0.0005303379,0.003227765,0.0007532801,0.001094349],"domain_scores_gemma":[0.9955186,0.0001124315,0.0003523213,0.003197894,0.000417464,0.0004012836],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.00001292643,0.00003321134,0.000009180299,0.0000793982,0.00001870481,0.0001824833,0.002982125,0.9087427,0.06506251,0.02241867,0.00001142196,0.0004466322],"study_design_scores_gemma":[0.0002547543,0.00003221198,0.000008040528,0.0004924475,0.00004154535,0.000006020413,0.000746389,0.9762566,0.007622072,0.01388435,0.000001612342,0.0006539737],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.5175138,0.00002005047,0.4801146,0.00004065566,0.001208939,0.0003781434,0.00002642406,0.0004143247,0.0002831216],"genre_scores_gemma":[0.9553631,0.000004541495,0.0430857,0.0001078618,0.0001755177,0.000003769774,0.00001603321,0.00006145349,0.001182011],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.4378494,"threshold_uncertainty_score":0.9995967,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.2573202568645652,"score_gpt":0.2837385523110563,"score_spread":0.02641829544649105,"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."}}