{"id":"W7115904749","doi":"10.5281/zenodo.17968427","title":"The ITI–gITI–IDI Measurement Framework for Budget–Outcome Tension With an Ontario EQAO Demonstration","year":2025,"lang":"","type":"preprint","venue":"Zenodo (CERN European Organization for Nuclear Research)","topic":"Public Policy and Administration Research","field":"Social Sciences","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"","keywords":"Ranking (information retrieval); Politics; Tension (geology); Attribution; Performance measurement; Term (time)","routes":{"ca_aff":false,"ca_fund":false,"ca_venue":false,"about_ca":true,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":["metaepi_narrow","sts","scholarly_communication","insufficient_payload"],"consensus_categories":[],"category_scores_codex":[0.009692245,0.0005830103,0.0005418874,0.0004038694,0.02319972,0.008446215,0.003926735,0.0006164752,0.00474887],"category_scores_gemma":[0.006176841,0.0004939614,0.0002599157,0.001265518,0.001492588,0.00082353,0.002029824,0.00206624,0.0007520093],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001872988,"about_ca_system_score_gemma":0.0008284385,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.005065974,"about_ca_topic_score_gemma":0.004132979,"domain_scores_codex":[0.9907633,0.002363628,0.00114943,0.001504537,0.002661226,0.001557886],"domain_scores_gemma":[0.990818,0.0004499702,0.0006725249,0.001512831,0.005686887,0.0008598289],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"not_applicable","study_design_scores_codex":[0.002399166,0.001514267,0.0001217577,0.0006547403,0.0007459118,0.0000182126,0.02649201,0.0004711811,0.0004193117,0.6920548,0.07737528,0.1977334],"study_design_scores_gemma":[0.0008044611,0.001124296,0.001178192,0.0003642752,0.0001034014,0.00002037674,0.005151209,0.001005438,0.0001279141,0.01224576,0.9772745,0.0006001998],"study_design_candidate":"not_applicable","study_design_consensus":null,"genre_codex":"other","genre_gemma":"empirical","genre_scores_codex":[0.02230351,0.0004703199,0.2610604,0.07575548,0.002600963,0.01833684,0.003039463,0.002653131,0.6137799],"genre_scores_gemma":[0.9645641,0.0007216149,0.004823943,0.001158634,0.001615562,0.000006907734,0.00358404,0.001687636,0.02183757],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.9422606,"threshold_uncertainty_score":0.9997512,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1400363320667955,"score_gpt":0.3525511717515503,"score_spread":0.2125148396847548,"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."}}