{"id":"W4389518977","doi":"10.18653/v1/2023.findings-emnlp.380","title":"HoneyBee: Progressive Instruction Finetuning of Large Language Models for Materials Science","year":2023,"lang":"en","type":"article","venue":"","topic":"Machine Learning in Materials Science","field":"Materials Science","cited_by":16,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Waterloo; Université de Montréal","funders":"Natural Sciences and Engineering Research Council of Canada; Canadian Institute for Advanced Research","keywords":"Computer science; Benchmark (surveying); Construct (python library); Trustworthiness; Language model; Process (computing); Soundness; Code (set theory); Quality (philosophy); Artificial intelligence; Machine learning; Data science; Programming language","routes":{"ca_aff":true,"ca_fund":true,"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":["insufficient_payload"],"consensus_categories":[],"category_scores_codex":[0.003315615,0.0001617677,0.0002893344,0.0003158315,0.0003672918,0.0002040395,0.0007263448,0.00005969765,0.001335153],"category_scores_gemma":[0.000660024,0.000133077,0.00003955965,0.0009487711,0.0004511137,0.0008135524,0.0003510302,0.00004812202,0.0001371986],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00003883319,"about_ca_system_score_gemma":0.0001527221,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00008745413,"about_ca_topic_score_gemma":0.000004690186,"domain_scores_codex":[0.9977281,0.00008013689,0.0004347345,0.0005270331,0.0005834291,0.0006465556],"domain_scores_gemma":[0.9988389,0.0001062984,0.0003124388,0.000412382,0.0002403221,0.00008967215],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"bench_or_experimental","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.00003270456,0.00001743236,0.00004162522,0.00009366793,0.000001357501,0.000002442731,0.001134454,0.00302906,0.9827163,0.01243914,0.0001966069,0.0002952697],"study_design_scores_gemma":[0.0004518232,0.00008713722,0.0005512996,0.00005097243,0.000005038535,0.000009474377,0.0006912353,0.03785342,0.9575345,0.002527809,0.00006777591,0.000169527],"study_design_candidate":"bench_or_experimental","study_design_consensus":"bench_or_experimental","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9892125,0.00001655226,0.007606727,0.0001173303,0.001203405,0.0004979905,0.0001367202,0.0004071109,0.0008016648],"genre_scores_gemma":[0.9730396,0.000002610449,0.02618725,0.00005125191,0.0001220477,0.00009699773,0.00002340665,0.0000216954,0.000455192],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.03482436,"threshold_uncertainty_score":0.9995778,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01707591023755806,"score_gpt":0.319318363640433,"score_spread":0.302242453402875,"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."}}