{"id":"W7082255705","doi":"10.48448/tdcz-8t38","title":"Rethinking Full Finetuning from Pretraining Checkpoints in Active Learning for African Languages","year":2025,"lang":"en","type":"other","venue":"Underline Science Inc.","topic":"Geochemistry and Geologic Mapping","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"McGill University; Mila - Quebec Artificial Intelligence Institute","funders":"","keywords":"Annotation; Key (lock); Active learning (machine learning); Training set; Language acquisition; Hybrid learning","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":[0.002327292,0.001270117,0.0008398384,0.0006521437,0.0006449896,0.001415255,0.002607044,0.00125023,0.00271859],"category_scores_gemma":[0.007363922,0.0005621451,0.0005944128,0.0005296314,0.001077273,0.002951031,0.002316791,0.002301096,0.001042279],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0006534272,"about_ca_system_score_gemma":0.001216814,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.008906583,"about_ca_topic_score_gemma":0.01660861,"domain_scores_codex":[0.9992501,0.0002602882,0.00004344418,0.0002528429,0.00008568878,0.0001075775],"domain_scores_gemma":[0.9967114,0.00227433,0.0001519684,0.000431769,0.0003100658,0.0001204235],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"not_applicable","study_design_scores_codex":[0.0005422623,0.000474251,0.008546147,0.0002172775,0.000157279,0.0001493426,0.0007198834,0.359521,0.02111218,0.002492283,0.002665902,0.6034022],"study_design_scores_gemma":[0.00003447137,0.0001395026,0.0009431473,0.00002539036,0.00002784252,0.00004821085,0.000205032,0.9832482,0.009843416,0.00388747,0.001576182,0.00002097811],"study_design_candidate":"not_applicable","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.2345981,0.0009819572,0.7466093,0.0007603584,0.0001532192,0.0002021152,0.0002874206,0.01152136,0.004886272],"genre_scores_gemma":[0.8366696,0.0001653808,0.1588935,0.0004134901,0.00003878663,0.0002064183,0.0004728208,0.0005489642,0.002591004],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.008906583,"threshold_uncertainty_score":0.01770943,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02228147574520321,"score_gpt":0.2782794241578435,"score_spread":0.2559979484126403,"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."}}