{"id":"W4389518858","doi":"10.18653/v1/2023.findings-emnlp.674","title":"Balaur: Language Model Pretraining with Lexical Semantic Relations","year":2023,"lang":"en","type":"article","venue":"","topic":"Topic Modeling","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"McGill University","funders":"Samsung; Nvidia","keywords":"Computer science; Natural language processing; Inference; Artificial intelligence; Generalization; Transformer; Set (abstract data type); Language model; Meaning (existential); Interface (matter); Programming language; Psychology","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":[],"consensus_categories":[],"category_scores_codex":[0.0001474339,0.00006145851,0.00006582923,0.00007321862,0.00007727319,0.00006178927,0.0002882689,0.00003213457,0.00002135567],"category_scores_gemma":[0.00002124232,0.00004769038,0.00002066048,0.000361058,0.00001382047,0.0002479075,0.0001156111,0.00009516034,0.0001759508],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00001276762,"about_ca_system_score_gemma":0.00004892697,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00001649018,"about_ca_topic_score_gemma":0.00002010635,"domain_scores_codex":[0.9992872,0.00001314001,0.00009804479,0.0002290691,0.0001822952,0.0001902554],"domain_scores_gemma":[0.9995027,0.00004792716,0.00001671483,0.0003623546,0.00001759695,0.00005266041],"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":[0.000002055284,0.00002007581,0.001874631,0.00001277507,0.00001779399,0.00005778991,0.01028105,0.4055216,0.0005584706,0.5626479,0.001793632,0.01721224],"study_design_scores_gemma":[0.00009618385,0.00001165202,0.0004176146,0.00001090835,0.000002061854,0.000006199501,0.0001326698,0.9964129,0.00005958178,0.002750273,0.00002460415,0.00007534672],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.07684215,0.00000723002,0.8991132,0.001288165,0.00003726038,0.00005553737,3.626427e-7,0.0006571017,0.02199898],"genre_scores_gemma":[0.7991173,5.605275e-7,0.1943886,0.0001470361,0.00001963586,0.000005379377,0.000001395855,0.000005295344,0.006314777],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.7222752,"threshold_uncertainty_score":0.226155,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02994062432139795,"score_gpt":0.2649370903985889,"score_spread":0.2349964660771909,"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."}}