{"id":"W6910292159","doi":"10.48448/17rh-v293","title":"Composable Sparse Fine-Tuning for Cross-Lingual Transfer","year":2022,"lang":"en","type":"other","venue":"Underline Science Inc.","topic":"","field":"","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Mila - Quebec Artificial Intelligence Institute","funders":"","keywords":"Inference; Set (abstract data type); Transfer (computing); Margin (machine learning); Task (project management); Language model; Forgetting; Simple (philosophy)","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","insufficient_payload"],"consensus_categories":["insufficient_payload"],"category_scores_codex":[0.002340842,0.0006469388,0.0006841015,0.001465258,0.001104677,0.0004767995,0.002303612,0.0002399632,0.01993545],"category_scores_gemma":[0.0002757173,0.0006316489,0.0002053616,0.001988634,0.002426415,0.0003205556,0.0003751129,0.0006732521,0.001428694],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0005992286,"about_ca_system_score_gemma":0.001735102,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0003577735,"about_ca_topic_score_gemma":0.001105502,"domain_scores_codex":[0.9947541,0.00007260549,0.00057719,0.001532674,0.001650314,0.001413142],"domain_scores_gemma":[0.9979373,0.0001988289,0.0002220695,0.00106388,0.0002430943,0.0003348374],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"not_applicable","study_design_gemma":"not_applicable","study_design_scores_codex":[0.0002626364,0.0008214066,0.001444716,0.0003814154,0.0001840982,0.00009265193,0.00104242,0.009998875,0.02849839,0.01806092,0.9322472,0.006965268],"study_design_scores_gemma":[0.00153942,0.0002374165,0.00002426628,0.00009426385,0.00007470423,0.0000295604,0.0001918799,0.0311574,0.000951617,0.0004842617,0.9642935,0.0009216646],"study_design_candidate":"not_applicable","study_design_consensus":"not_applicable","genre_codex":"other","genre_gemma":"other","genre_scores_codex":[0.005658827,0.001234813,0.01780672,0.0003580706,0.004545221,0.004305642,0.006921364,0.003089093,0.9560803],"genre_scores_gemma":[0.04670338,0.00001938034,0.03658507,0.0005935908,0.001759686,0.0003456621,0.001649544,0.003319667,0.909024],"genre_candidate":"other","genre_consensus":"other","teacher_disagreement_score":0.04705623,"threshold_uncertainty_score":0.9996135,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.05051522559580456,"score_gpt":0.3485111589177882,"score_spread":0.2979959333219837,"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."}}