{"id":"W7077887124","doi":"10.48448/rdjn-9j55","title":"ACCORD: Closing the Commonsense Measurability Gap","year":2025,"lang":"en","type":"other","venue":"Underline Science Inc.","topic":"Geochemistry and Geologic Mapping","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Dalhousie University","funders":"","keywords":"Commonsense reasoning; Suite; Benchmark (surveying); Closing (real estate); Automated reasoning; Commonsense knowledge; Code (set theory)","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":[0.008781229,0.002270955,0.001033878,0.002253589,0.001346905,0.00434891,0.004223903,0.002097697,0.0069955],"category_scores_gemma":[0.05226301,0.0008545919,0.001978832,0.002053514,0.003947662,0.007483006,0.006282077,0.004806875,0.001479922],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.002254962,"about_ca_system_score_gemma":0.004572143,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.007783717,"about_ca_topic_score_gemma":0.01210086,"domain_scores_codex":[0.9859197,0.006649089,0.0007693995,0.001435412,0.004505463,0.0007209895],"domain_scores_gemma":[0.962869,0.02387741,0.001189641,0.008129489,0.002965365,0.0009690962],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.001385284,0.001221163,0.01043363,0.00248479,0.0003154911,0.0006139187,0.001075325,0.2962851,0.01203619,0.3587761,0.1018266,0.2135464],"study_design_scores_gemma":[0.0002745682,0.0003080803,0.001015654,0.00025781,0.00009787537,0.0002235212,0.0002451922,0.6798874,0.01560578,0.2607598,0.04123405,0.00009029386],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.1762483,0.004676355,0.6803052,0.008377653,0.001229802,0.0007226001,0.008149243,0.05431431,0.06597659],"genre_scores_gemma":[0.6181402,0.0008464886,0.3553353,0.001763695,0.0002315779,0.000538383,0.01028142,0.007498107,0.005364776],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.008781229,"threshold_uncertainty_score":0.04644012,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03445118359345169,"score_gpt":0.2781933986903419,"score_spread":0.2437422150968902,"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."}}