{"id":"W4385932377","doi":"10.1101/2023.08.15.553415","title":"Effect of Tokenization on Transformers for Biological Sequences","year":2023,"lang":"en","type":"preprint","venue":"bioRxiv (Cold Spring Harbor Laboratory)","topic":"Genomics and Phylogenetic Studies","field":"Biochemistry, Genetics and Molecular Biology","cited_by":5,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Tel Aviv University; Azrieli Foundation; Israel Science Foundation","keywords":"Lexical analysis; Computer science; Artificial intelligence; Natural language processing; Biological data; DNA sequencing; Inference; Machine learning; Biology; Bioinformatics; DNA; Genetics","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.006564141,0.001544098,0.001011697,0.001295829,0.0007440921,0.001930436,0.002409986,0.001214933,0.00339394],"category_scores_gemma":[0.03724114,0.000509885,0.000871668,0.001256228,0.001687084,0.005708305,0.001836912,0.002259743,0.001147241],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001692162,"about_ca_system_score_gemma":0.002999963,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.004506891,"about_ca_topic_score_gemma":0.004593495,"domain_scores_codex":[0.9958604,0.001426431,0.0005656559,0.001050408,0.000585429,0.0005116689],"domain_scores_gemma":[0.9575783,0.03159148,0.002110086,0.0050324,0.002701849,0.0009858897],"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":[0.01011712,0.001472942,0.02106549,0.001146733,0.000481974,0.0009467548,0.0005480985,0.4306866,0.06001119,0.0159704,0.01168022,0.4458725],"study_design_scores_gemma":[0.0002980032,0.001062009,0.001543632,0.00005423895,0.0001969495,0.000208268,0.0003124765,0.9011298,0.08271358,0.009767454,0.00265258,0.00006102298],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"methods","genre_scores_codex":[0.8634923,0.001214651,0.1158391,0.001146662,0.0007441556,0.0002239373,0.0009793151,0.01307205,0.003287859],"genre_scores_gemma":[0.9314769,0.0002630481,0.06416854,0.0003127243,0.00007559214,0.00009927723,0.001604876,0.0006885836,0.001310538],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.006564141,"threshold_uncertainty_score":0.03471488,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01992295486608473,"score_gpt":0.2485272778004797,"score_spread":0.228604322934395,"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."}}