{"id":"W4416186693","doi":"10.48550/arxiv.2509.08105","title":"MERLIN: Multi-Stage Curriculum Alignment for Multilingual Encoder-LLM Integration in Cross-Lingual Reasoning","year":2025,"lang":"en","type":"preprint","venue":"ArXiv.org","topic":"Natural Language Processing Techniques","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Natural Sciences and Engineering Research Council of Canada; Institut de Valorisation des Données; Canada First Research Excellence Fund","keywords":"Benchmark (surveying); Curriculum; Set (abstract data type); Data set; Training set; Computational linguistics; Language acquisition","routes":{"ca_aff":false,"ca_fund":true,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.002397562,0.002046314,0.001053402,0.001620594,0.001074703,0.002053992,0.003468715,0.001840216,0.01991701],"category_scores_gemma":[0.007050965,0.001339645,0.00127583,0.00144056,0.0007220735,0.005660343,0.004403268,0.003525664,0.00998744],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001333115,"about_ca_system_score_gemma":0.002462219,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.007083803,"about_ca_topic_score_gemma":0.02315241,"domain_scores_codex":[0.9982637,0.0005818315,0.0001027286,0.0005979972,0.0002961725,0.0001576533],"domain_scores_gemma":[0.998478,0.0006007418,0.00007093934,0.0005059562,0.0002591819,0.00008527379],"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":[0.0003963059,0.0003457657,0.00177102,0.0003684084,0.0002309486,0.0002101626,0.0003970678,0.04257741,0.01123783,0.01300202,0.04382105,0.885642],"study_design_scores_gemma":[0.0001475009,0.0001645843,0.0006924034,0.0000710687,0.00008693588,0.0001806866,0.0002724934,0.9025561,0.02730167,0.0464866,0.02198218,0.00005772678],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.02002533,0.0007696599,0.8794284,0.0004312056,0.0002156489,0.0001698239,0.001369822,0.091683,0.005907077],"genre_scores_gemma":[0.2085167,0.0002844365,0.7713386,0.000491257,0.00009602265,0.0002645226,0.006488078,0.004285492,0.008234942],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.01991701,"threshold_uncertainty_score":0.06662905,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.05660995828063751,"score_gpt":0.3846209979870951,"score_spread":0.3280110397064576,"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."}}