{"id":"W4416394486","doi":"10.48550/arxiv.2510.07500","title":"Black-Box Detection of LLM-Generated Text Using Generalized Jensen-Shannon Divergence","year":2025,"lang":"en","type":"preprint","venue":"ArXiv.org","topic":"Authorship Attribution and Profiling","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Government of Canada; Canadian Institute for Advanced Research","keywords":"Divergence (linguistics); Detector; Asymptotic distribution; Discretization; Pattern recognition (psychology); Matrix (chemical analysis); Statistical hypothesis testing","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":["metaresearch"],"consensus_categories":[],"category_scores_codex":[0.00705437,0.0009372825,0.001237144,0.003034218,0.0008885315,0.002902366,0.001832145,0.001930193,0.002009328],"category_scores_gemma":[0.04332181,0.0005106012,0.0006258982,0.001948355,0.001924989,0.003809572,0.002836554,0.002072021,0.001303334],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001316663,"about_ca_system_score_gemma":0.001232671,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.001292603,"about_ca_topic_score_gemma":0.001787258,"domain_scores_codex":[0.9949397,0.002164179,0.0003112222,0.001142205,0.001159746,0.0002830143],"domain_scores_gemma":[0.9644793,0.02682335,0.002371654,0.003642724,0.002007749,0.0006751568],"domain_codex":null,"domain_gemma":"methods","domain_candidate":"methods","domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.001336784,0.0003127694,0.0433296,0.0008370215,0.0002373307,0.001482139,0.002171817,0.3241498,0.03404444,0.0894168,0.01346229,0.4892192],"study_design_scores_gemma":[0.00001830667,0.00007038702,0.002561551,0.00003396207,0.00001015559,0.0001883259,0.00009043028,0.9436961,0.01201335,0.03973623,0.001541046,0.000040217],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.1185873,0.0007218606,0.8729765,0.0005798994,0.0001031929,0.00008228806,0.0008692064,0.003834711,0.002245096],"genre_scores_gemma":[0.8231311,0.0001912964,0.1713095,0.0002566341,0.0001212103,0.000151616,0.002152824,0.0004960797,0.002189754],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.9929456,"threshold_uncertainty_score":0.03730756,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.09692107733937994,"score_gpt":0.3169443602513833,"score_spread":0.2200232829120033,"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."}}