{"id":"W4414954382","doi":"10.2172/2997408","title":"Summer Internship Report: ARA2 Benchmarking","year":2025,"lang":"en","type":"report","venue":"","topic":"Human Resource Development and Performance Evaluation","field":"Psychology","cited_by":0,"is_retracted":false,"has_abstract":false,"ca_institutions":"University of Waterloo","funders":"Lawrence Livermore National Laboratory; U.S. Department of Energy","keywords":"","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":[0.007827164,0.001281838,0.000635329,0.002048641,0.002215758,0.003760446,0.001731179,0.002633598,0.1758973],"category_scores_gemma":[0.005418554,0.0004396838,0.0005691496,0.001520365,0.0003770744,0.001163346,0.001836508,0.002441566,0.08576786],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.002641118,"about_ca_system_score_gemma":0.01292483,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.04922395,"about_ca_topic_score_gemma":0.07808675,"domain_scores_codex":[0.9961509,0.0005788181,0.0001148149,0.0002975301,0.002014075,0.0008438485],"domain_scores_gemma":[0.9892076,0.000736045,0.0002585336,0.0005115395,0.005501357,0.003785007],"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":[0.0003731746,0.0003698354,0.001641007,0.0000301435,0.000005901099,0.00003847663,0.00002118289,0.0002866153,0.0004543027,0.0005214083,0.9837974,0.01246062],"study_design_scores_gemma":[0.0003491243,0.0008702874,0.04995187,0.00005918966,0.00001671814,0.00007585093,0.0004650454,0.002643876,0.003469784,0.0007393192,0.9413095,0.00004946208],"study_design_candidate":"not_applicable","study_design_consensus":"not_applicable","genre_codex":"other","genre_gemma":"other","genre_scores_codex":[0.02906602,0.0007495477,0.00518184,0.02956847,0.009225052,0.00341235,0.2458946,0.006201944,0.6707002],"genre_scores_gemma":[0.02166544,0.0002314055,0.001935425,0.001812943,0.0005748772,0.0008399134,0.06341747,0.0009353159,0.9085872],"genre_candidate":"other","genre_consensus":"other","teacher_disagreement_score":0.1758973,"threshold_uncertainty_score":0.5884351,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1733975394905388,"score_gpt":0.445259909225525,"score_spread":0.2718623697349862,"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."}}