{"id":"W4405096276","doi":"10.48550/arxiv.2408.11646","title":"Mathematical Information Retrieval: Search and Question Answering","year":2024,"lang":"en","type":"preprint","venue":"arXiv (Cornell University)","topic":"Intelligent Tutoring Systems and Adaptive Learning","field":"Computer Science","cited_by":1,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Queen's University; Alfred P. Sloan Foundation; National Science Foundation","keywords":"Question answering; Information retrieval; Computer science","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":[],"consensus_categories":[],"category_scores_codex":[0.000568613,0.0001796837,0.0001878689,0.0002706767,0.0001066447,0.0004257084,0.0004596605,0.0001883812,0.000007651016],"category_scores_gemma":[0.00004684955,0.0001910807,0.00007698352,0.0002914054,0.00003951037,0.0005265813,0.001690263,0.0007053445,0.0002710216],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001651365,"about_ca_system_score_gemma":0.00008544427,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0000818345,"about_ca_topic_score_gemma":0.000001499133,"domain_scores_codex":[0.9989406,0.00009516961,0.000190444,0.0004416357,0.0001158203,0.0002163694],"domain_scores_gemma":[0.9992318,0.00006223855,0.0000740016,0.0004062775,0.0001306292,0.00009502037],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.000006328624,0.000007034619,0.0001141808,0.0004224159,0.00002829601,0.00005612249,0.0009472729,0.04726174,0.00002736299,0.9502516,0.00001695131,0.0008606563],"study_design_scores_gemma":[0.00008531682,0.00004034725,0.0002307081,0.0007187557,0.00002407021,0.00001631101,0.0001910041,0.9310273,0.0002423713,0.06535068,0.001792932,0.0002802147],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.2263156,0.00006837043,0.7703373,0.00008023194,0.0004722747,0.0001742457,0.000002415624,0.0002921278,0.002257428],"genre_scores_gemma":[0.9948936,0.00005630891,0.001583388,0.00001686869,0.00007569374,3.062044e-7,0.000004056364,0.000009358321,0.003360483],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.8849009,"threshold_uncertainty_score":0.7792047,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.06316537991141445,"score_gpt":0.2086274525411095,"score_spread":0.145462072629695,"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."}}