{"id":"W7104183057","doi":"","title":"LLMs as Judges: Toward The Automatic Review of GSN-compliant Assurance Cases","year":2025,"lang":"","type":"article","venue":"ArXiv.org","topic":"Safety Systems Engineering in Autonomy","field":"Engineering","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Canada First Research Excellence Fund","keywords":"Leverage (statistics); Quality assurance; Harm; Task (project management); Obstacle; Risk assessment","routes":{"ca_aff":false,"ca_fund":true,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":["metaepi_narrow","insufficient_payload"],"consensus_categories":[],"category_scores_codex":[0.001038731,0.0007320484,0.00138058,0.0002018258,0.0001400171,0.00004951425,0.001217379,0.0002354771,0.0005990193],"category_scores_gemma":[0.001629451,0.0006207682,0.0005070632,0.001315062,0.0001864945,0.000221589,0.0002343875,0.0006460028,0.001019253],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0004070311,"about_ca_system_score_gemma":0.0003590461,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0001987945,"about_ca_topic_score_gemma":0.00000801023,"domain_scores_codex":[0.9960259,0.0002063075,0.001935343,0.0005979011,0.0004604011,0.0007741841],"domain_scores_gemma":[0.9958444,0.0014265,0.0003686614,0.001997952,0.0002210148,0.0001415211],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"systematic_review","study_design_gemma":"observational","study_design_scores_codex":[0.00004638115,0.0005779163,0.1745889,0.6463093,0.006343809,0.0006181033,0.005558033,0.05631888,0.006767801,0.007537029,0.0614121,0.03392183],"study_design_scores_gemma":[0.001506679,0.0002348186,0.5072103,0.2219655,0.00150212,0.0004574209,0.0009097878,0.05175503,0.006364212,0.0001344916,0.2058488,0.002110855],"study_design_candidate":"systematic_review","study_design_consensus":null,"genre_codex":"review","genre_gemma":"empirical","genre_scores_codex":[0.2711058,0.6595979,0.01362226,0.01291746,0.009940693,0.004663812,0.0001011481,0.001517161,0.02653382],"genre_scores_gemma":[0.976248,0.02079754,0.0005041896,0.0008998588,0.0001950062,0.000256299,0.000009709011,0.0001126696,0.0009766989],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.7051423,"threshold_uncertainty_score":0.9997585,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03308860752920712,"score_gpt":0.2738435258553596,"score_spread":0.2407549183261525,"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."}}