{"id":"W4416023900","doi":"10.48550/arxiv.2511.04053","title":"Interpreting Multi-Attribute Confounding through Numerical Attributes in Large Language Models","year":2025,"lang":"","type":"preprint","venue":"ArXiv.org","topic":"Explainable Artificial Intelligence (XAI)","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Core Research for Evolutional Science and Technology; Institute for Catastrophic Loss Reduction","keywords":"Context (archaeology); Vulnerability (computing); Linear subspace; ENCODE; Language model; Hierarchy; Correlation; Numerical modeling","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":[0.01470369,0.0008791852,0.0009702935,0.001116361,0.00101597,0.005093766,0.001718679,0.001206752,0.003388828],"category_scores_gemma":[0.1148045,0.0008710602,0.001470858,0.0008349375,0.006056861,0.01228381,0.007397589,0.003433798,0.0002189674],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001614451,"about_ca_system_score_gemma":0.00135761,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002337795,"about_ca_topic_score_gemma":0.001815292,"domain_scores_codex":[0.99103,0.006039481,0.0003831044,0.00140238,0.000808095,0.0003369875],"domain_scores_gemma":[0.9077519,0.07300573,0.007267037,0.009568943,0.001417757,0.0009885906],"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":[0.0009376075,0.0002197014,0.06343257,0.0004578638,0.0005980763,0.0009207801,0.01001609,0.2063259,0.01445,0.6375945,0.001202917,0.06384395],"study_design_scores_gemma":[0.00003269606,0.00005578776,0.003970995,0.00003429094,0.00007459684,0.00009304111,0.0004597489,0.3130206,0.002838188,0.678762,0.0006050688,0.00005305142],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.4694205,0.0001301281,0.5242761,0.00164515,0.00004078936,0.00006455965,0.0001883701,0.000612343,0.003622064],"genre_scores_gemma":[0.963412,0.00003932452,0.03589846,0.00012786,0.00001975,0.00005196637,0.0001017478,0.00009553887,0.0002532454],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.01470369,"threshold_uncertainty_score":0.07776147,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1102617408806471,"score_gpt":0.3656238299806145,"score_spread":0.2553620890999674,"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."}}