{"id":"W4393749247","doi":"10.5281/zenodo.6979991","title":"GENEA Challenge 2022 objective evaluation data","year":2022,"lang":"en","type":"dataset","venue":"Zenodo (CERN European Organization for Nuclear Research)","topic":"Advanced Data Processing Techniques","field":"Engineering","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Electronic Arts (Canada)","funders":"","keywords":"Computer science; Data science","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":[0.009346053,0.004299557,0.002491927,0.003750794,0.00122624,0.003151851,0.004049666,0.002958855,0.05067216],"category_scores_gemma":[0.02428349,0.0007986104,0.001952397,0.002225534,0.000824179,0.002516091,0.004233161,0.002381901,0.05688483],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001852155,"about_ca_system_score_gemma":0.002242329,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.01460427,"about_ca_topic_score_gemma":0.02359313,"domain_scores_codex":[0.9865919,0.004272124,0.0009964085,0.001690226,0.005679601,0.000769847],"domain_scores_gemma":[0.9816426,0.003429608,0.0007932531,0.003364393,0.009364969,0.001405139],"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":[0.001583262,0.001072958,0.003315272,0.00221486,0.0002783797,0.0001671008,0.0001959428,0.008804571,0.00329047,0.001130601,0.8853442,0.09260236],"study_design_scores_gemma":[0.002303893,0.002878232,0.0486907,0.001149475,0.0003196148,0.0006612235,0.0008572588,0.07681769,0.01343869,0.00559177,0.8466956,0.0005959221],"study_design_candidate":"not_applicable","study_design_consensus":"not_applicable","genre_codex":"dataset","genre_gemma":"dataset","genre_scores_codex":[0.04439103,0.004860485,0.02922192,0.001379252,0.002262803,0.006383024,0.7882074,0.04218732,0.08110681],"genre_scores_gemma":[0.0500651,0.0004603697,0.02295443,0.0006029648,0.0002131917,0.00737885,0.8887298,0.003958216,0.02563709],"genre_candidate":"dataset","genre_consensus":"dataset","teacher_disagreement_score":0.05067216,"threshold_uncertainty_score":0.1695153,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.08138887783894873,"score_gpt":0.3070438007439275,"score_spread":0.2256549229049788,"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."}}