{"id":"W6910435407","doi":"10.48448/w0ps-0t42","title":"BPE Gets Picky: Efficient Vocabulary Refinement During Tokenizer Training","year":2024,"lang":"en","type":"other","venue":"Underline Science Inc.","topic":"","field":"","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Thomson Reuters (Canada)","funders":"","keywords":"Vocabulary; Downstream (manufacturing); Encoding (memory); Training (meteorology); Simple (philosophy); Harm; Compromise","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.000869244,0.001556385,0.0007667658,0.0009943742,0.0006312176,0.001656355,0.002192916,0.0009840154,0.02477173],"category_scores_gemma":[0.005531937,0.0007074943,0.0005677506,0.0009309727,0.0007500162,0.003916069,0.00268954,0.002172374,0.02089002],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0005418572,"about_ca_system_score_gemma":0.001326245,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002722133,"about_ca_topic_score_gemma":0.00405574,"domain_scores_codex":[0.9990863,0.0001424722,0.00008462322,0.00029056,0.00029135,0.0001046742],"domain_scores_gemma":[0.9986098,0.0004150062,0.00008501024,0.0005273745,0.0002853607,0.00007751412],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0007717604,0.000138298,0.001406132,0.0003191987,0.00008687357,0.0004169371,0.0002556943,0.01691096,0.06905057,0.01279304,0.04380986,0.8540407],"study_design_scores_gemma":[0.0001729419,0.0002299632,0.001220623,0.00007533743,0.00007158878,0.0007430682,0.0001844365,0.4880749,0.3909691,0.02773067,0.09039307,0.0001343463],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.01388521,0.0003202494,0.9257613,0.0003417445,0.0003433087,0.0001668973,0.001484428,0.05112137,0.006575519],"genre_scores_gemma":[0.1820638,0.0003268103,0.7651492,0.0004718384,0.0001874743,0.0003159606,0.008071965,0.01125064,0.03216223],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.02477173,"threshold_uncertainty_score":0.08286965,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03139523454099037,"score_gpt":0.2981416731004584,"score_spread":0.266746438559468,"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."}}