{"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":"codex-gemma-dda1882f352a","candidate_categories":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.001047337,0.000308492,0.000388363,0.0006029747,0.0002664432,0.000229174,0.0019788,0.0002770335,0.0001754023],"category_scores_gemma":[0.001951061,0.0003065358,0.00006633304,0.001190303,0.0003045195,0.0002493856,0.0007196263,0.0007072861,0.00001035053],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001400943,"about_ca_system_score_gemma":0.0006469247,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0006788197,"about_ca_topic_score_gemma":0.0004809589,"domain_scores_codex":[0.9974263,0.00006822313,0.0003151101,0.001112716,0.0004254993,0.0006521233],"domain_scores_gemma":[0.9984004,0.0004450541,0.0003526704,0.0005677764,0.0001384378,0.00009559955],"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":[0.00009675008,0.0003228642,0.00159536,0.0007698756,0.0002571898,0.0002506524,0.04791448,0.01004598,0.02596445,0.05813685,0.0277742,0.8268713],"study_design_scores_gemma":[0.001642123,0.0002454081,0.0005384352,0.003341394,0.00004350616,0.00001331785,0.005317141,0.7408887,0.006832322,0.05814777,0.1811971,0.001792812],"study_design_candidate":"design_other","study_design_consensus":null,"genre_codex":"other","genre_gemma":"empirical","genre_scores_codex":[0.0002746901,0.0002841528,0.2453411,0.001756224,0.0004543714,0.0004193686,0.000021589,0.0004923015,0.7509562],"genre_scores_gemma":[0.359228,0.00003284524,0.3203017,0.0004296786,0.0005601746,0.0000857259,0.0001259013,0.00006820234,0.3191679],"genre_candidate":"other","genre_consensus":null,"teacher_disagreement_score":0.8250785,"threshold_uncertainty_score":0.9999387,"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."}}