{"id":"W4405502038","doi":"10.2139/ssrn.5061381","title":"Beyond Vanilla Fine-Tuning: Leveraging Multistage, Multilingual, and Domain-Specific Methods for Low-Resource Machine Translation","year":2024,"lang":"en","type":"preprint","venue":"SSRN Electronic Journal","topic":"Natural Language Processing Techniques","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":false,"ca_institutions":"University of Toronto","funders":"","keywords":"Computer science; Machine translation; Fine-tuning; Translation (biology); Resource (disambiguation); Domain (mathematical analysis); Artificial intelligence; Mathematics; Chemistry; Particle physics; Physics; Computer network","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":[0.002652112,0.001545364,0.002010724,0.001269209,0.001149402,0.002750367,0.002356751,0.001920463,0.007772971],"category_scores_gemma":[0.008998176,0.000806472,0.001253616,0.001954936,0.001080386,0.004030874,0.003468984,0.002815616,0.005127598],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0005269899,"about_ca_system_score_gemma":0.001940697,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002480903,"about_ca_topic_score_gemma":0.008615821,"domain_scores_codex":[0.9973322,0.001184826,0.0001646907,0.0006429417,0.0004318584,0.0002434976],"domain_scores_gemma":[0.9959387,0.00231976,0.0001397789,0.001066738,0.0004119068,0.0001232705],"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":[0.000915026,0.0005247543,0.001229586,0.0006883221,0.0003485792,0.0004180023,0.0003721656,0.08732976,0.04679329,0.02041984,0.01496736,0.8259933],"study_design_scores_gemma":[0.00009335285,0.0001616754,0.0004132283,0.00006923515,0.0001220231,0.0002891654,0.0001127904,0.89966,0.02180787,0.06573874,0.0114556,0.00007632635],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.02312028,0.00173854,0.9560596,0.000436343,0.0002569759,0.0001063279,0.000248776,0.01230499,0.005728269],"genre_scores_gemma":[0.3826793,0.0008143943,0.601981,0.000786002,0.0002937697,0.0001803587,0.00156602,0.002976815,0.00872222],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.007772971,"threshold_uncertainty_score":0.02600324,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02022335911312361,"score_gpt":0.3356685133325591,"score_spread":0.3154451542194355,"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."}}