{"id":"W2915258134","doi":"","title":"Semeval-2012 Task 8: Cross-lingual Textual Entailment for Content Synchronization","year":2012,"lang":"en","type":"article","venue":"","topic":"Natural Language Processing Techniques","field":"Computer Science","cited_by":28,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of British Columbia","funders":"","keywords":"Computer science; Textual entailment; Task (project management); Natural language processing; SemEval; Logical consequence; Inference; Synchronization (alternating current); Artificial intelligence; Process (computing); Programming language","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":[0.000456959,0.0001508164,0.0001300665,0.00007466662,0.0001529006,0.0002452114,0.000584332,0.00008563529,0.00004982694],"category_scores_gemma":[0.0001144174,0.0001187052,0.00006356539,0.0001642535,0.00004223022,0.001502615,0.000246399,0.00008839725,0.0000359161],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001317267,"about_ca_system_score_gemma":0.00004553818,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00002188106,"about_ca_topic_score_gemma":0.000004712685,"domain_scores_codex":[0.9987478,0.00002640059,0.000242814,0.0002698642,0.0002639455,0.0004492113],"domain_scores_gemma":[0.9991516,0.0000758513,0.0001028221,0.0003360743,0.0002135781,0.0001201163],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.00003712154,0.0004572242,0.007909318,0.0001426464,0.00005930987,0.000003824228,0.002385424,0.00001301675,0.03460911,0.5731849,0.009707572,0.3714905],"study_design_scores_gemma":[0.002084562,0.0005521227,0.00168708,0.0001089202,0.00004881886,0.00009537803,0.0002644866,0.03700658,0.9064523,0.01560178,0.03468848,0.001409511],"study_design_candidate":"bench_or_experimental","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.006150414,0.002553676,0.9889271,0.0003219621,0.0004158947,0.0004118212,0.000005182678,0.0007359216,0.0004779728],"genre_scores_gemma":[0.6612356,0.000003511754,0.33681,0.0005008491,0.0001987028,0.00004562078,0.0000126667,0.00001001405,0.001182992],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.8718432,"threshold_uncertainty_score":0.4840659,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03341947877365044,"score_gpt":0.3240185823605777,"score_spread":0.2905991035869273,"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."}}