{"id":"W7126460285","doi":"10.18653/v1/2022.wmt-1.8","title":"Test Set Sampling Affects System Rankings: Expanded Human Evaluation of WMT20 English-Inuktitut Systems","year":2022,"lang":"en","type":"article","venue":"","topic":"Psychometric Methodologies and Testing","field":"Decision Sciences","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"National Research Council Canada","funders":"","keywords":"Set (abstract data type); Sampling (signal processing); Test (biology); Data set; Test set","routes":{"ca_aff":true,"ca_fund":false,"ca_venue":false,"about_ca":true,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"metacan-v3-hybrid-931329e0061c","candidate_categories":["metaresearch"],"consensus_categories":[],"category_scores_codex":[0.02094933,0.001973822,0.001625218,0.002638582,0.002307684,0.002808033,0.001802501,0.001794798,0.003963111],"category_scores_gemma":[0.08584023,0.0005544064,0.0007459095,0.00290978,0.001639655,0.002005174,0.003351339,0.001608082,0.002257622],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001864275,"about_ca_system_score_gemma":0.001653779,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.02059444,"about_ca_topic_score_gemma":0.05080048,"domain_scores_codex":[0.9572628,0.0288032,0.002595787,0.00454755,0.005614955,0.001175619],"domain_scores_gemma":[0.9094551,0.06107118,0.002782126,0.01109875,0.01357313,0.002019767],"domain_codex":null,"domain_gemma":"evaluation","domain_candidate":"evaluation","domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"observational","study_design_scores_codex":[0.009482509,0.003318668,0.1261887,0.003561653,0.002613734,0.002094893,0.01503245,0.07360744,0.08500532,0.002615937,0.1219267,0.5545521],"study_design_scores_gemma":[0.002059866,0.006528223,0.3541065,0.0006036543,0.001218428,0.002596216,0.00938783,0.3933518,0.1263773,0.008020924,0.09470259,0.001046675],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9493235,0.00250623,0.02117881,0.0007563907,0.0006529132,0.0004848056,0.006648425,0.004519813,0.01392907],"genre_scores_gemma":[0.9484863,0.0002258114,0.02164333,0.0003916132,0.0001642547,0.000429432,0.02273726,0.001167926,0.00475405],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.9790507,"threshold_uncertainty_score":0.110792,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.6765852235668929,"score_gpt":0.5162689098489014,"score_spread":0.1603163137179915,"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."}}