{"id":"W7117339587","doi":"","title":"A Benchmark for Evaluating Outcome-Driven Constraint Violations in Autonomous AI Agents","year":2025,"lang":"","type":"article","venue":"ArXiv.org","topic":"Ethics and Social Impacts of AI","field":"Social Sciences","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Institut de Valorisation des Données; Natural Sciences and Engineering Research Council of Canada; Canada Research Chairs","keywords":"Benchmark (surveying); Task (project management); Key (lock); Constraint (computer-aided design); Software deployment; Incentive; Misconduct; Baseline (sea)","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.004116561,0.00149051,0.0005438143,0.001306261,0.0005024181,0.001293193,0.001498592,0.001092584,0.002084828],"category_scores_gemma":[0.01687906,0.0003117756,0.0007190788,0.0009571335,0.0008604271,0.001641316,0.001503854,0.001604195,0.0006296523],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001471936,"about_ca_system_score_gemma":0.001696383,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.01012442,"about_ca_topic_score_gemma":0.01313315,"domain_scores_codex":[0.9978211,0.0009486772,0.0001922747,0.0003301562,0.0005085308,0.0001993619],"domain_scores_gemma":[0.991688,0.004541729,0.0007662614,0.00134902,0.001204525,0.0004504577],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0005742402,0.0005147711,0.02653342,0.0006291632,0.0001921846,0.0001779146,0.0002666665,0.9013143,0.004266318,0.006277152,0.009885028,0.0493688],"study_design_scores_gemma":[0.00008484383,0.000630461,0.005460924,0.0001024841,0.0000336512,0.00004896136,0.0001903716,0.977321,0.004587745,0.006236805,0.0052683,0.00003439606],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"empirical","genre_gemma":"methods","genre_scores_codex":[0.875379,0.002862746,0.08822778,0.001470315,0.0003004373,0.0003433649,0.00589784,0.00723546,0.01828303],"genre_scores_gemma":[0.9320684,0.0004143803,0.05667896,0.000280402,0.00002635357,0.0002636151,0.008367377,0.0003737565,0.00152689],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.01012442,"threshold_uncertainty_score":0.02177072,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.191925008745504,"score_gpt":0.4866832160748377,"score_spread":0.2947582073293337,"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."}}