{"id":"W3206907172","doi":"10.18653/v1/2022.acl-long.125","title":"Composable Sparse Fine-Tuning for Cross-Lingual Transfer","year":2022,"lang":"en","type":"article","venue":"Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","topic":"Topic Modeling","field":"Computer Science","cited_by":62,"is_retracted":false,"has_abstract":true,"ca_institutions":"McGill University; Mila - Quebec Artificial Intelligence Institute","funders":"","keywords":"Computer science; Language model; Fine-tuning; Modular design; Overfitting; Artificial intelligence; Task (project management); Inference; Adapter (computing); Programming language; Computer hardware","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":[0.002126514,0.0002054646,0.0003405458,0.00009321197,0.001166862,0.0001283063,0.001739483,0.0000808891,0.000003609358],"category_scores_gemma":[0.006437561,0.0001764748,0.0003997594,0.0004307015,0.00007085304,0.0001085928,0.0005955502,0.0002874186,5.636962e-7],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0003047113,"about_ca_system_score_gemma":0.0002140652,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00002837413,"about_ca_topic_score_gemma":0.000003513341,"domain_scores_codex":[0.9973406,0.0000349206,0.0007757914,0.0004104088,0.001040047,0.0003982929],"domain_scores_gemma":[0.9949918,0.0009882052,0.0009633953,0.0001932087,0.002806414,0.00005692282],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0001120101,0.000167934,0.08859748,0.0003191032,0.0001811233,1.430814e-7,0.003534246,0.7896556,0.001083309,0.1141125,0.001824809,0.0004118214],"study_design_scores_gemma":[0.001803948,0.0002235227,0.009749539,0.000163382,0.0001500643,0.000004503639,0.0004811678,0.9357015,0.003102707,0.03250015,0.0156867,0.0004328129],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.861628,0.0002656407,0.1057421,0.005209234,0.01065427,0.004357468,0.001847369,0.0004095654,0.009886298],"genre_scores_gemma":[0.9640948,4.798006e-7,0.03381086,0.00021198,0.0003796327,0.00006970146,0.00001537425,0.00002889239,0.001388349],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.146046,"threshold_uncertainty_score":0.8974674,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01407345951011647,"score_gpt":0.2517387941339686,"score_spread":0.2376653346238521,"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."}}