{"id":"W7125792265","doi":"10.21428/594757db.d23b1499","title":"Enhancing Legal Text Entailment: Evaluating Model Architectures, Training Approaches, and Interpretability","year":2025,"lang":"en","type":"article","venue":"","topic":"Artificial Intelligence in Law","field":"Social Sciences","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Ottawa","funders":"","keywords":"Interpretability; Textual entailment; Context (archaeology); Logical consequence; Task (project management); Language model; Adaptation (eye); Focus (optics)","routes":{"ca_aff":true,"ca_fund":false,"ca_venue":false,"about_ca":true,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.002466428,0.0001308968,0.0001834431,0.0000808132,0.0006832864,0.0001807341,0.0002519215,0.00009597754,0.0002024509],"category_scores_gemma":[0.0009889087,0.0001233128,0.00006078766,0.0002332253,0.0005535517,0.0001705488,0.0001358063,0.0002395096,0.000006915109],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001523053,"about_ca_system_score_gemma":0.0003932369,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002683669,"about_ca_topic_score_gemma":0.0148749,"domain_scores_codex":[0.9982336,0.0002829898,0.0003503676,0.0004042861,0.0003268292,0.0004019526],"domain_scores_gemma":[0.9991762,0.0004254645,0.00005867534,0.0001903119,0.00005093586,0.00009843538],"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":[0.00003035145,0.00004246831,0.0006636544,0.0000267804,0.00002544524,5.909278e-7,0.08781546,0.004678912,0.002119351,0.5124815,0.00003092479,0.3920846],"study_design_scores_gemma":[0.0001446071,0.00006274134,0.0001070015,0.0001846918,0.00005659927,0.000001097319,0.109777,0.6838055,0.01625188,0.1856962,0.00347284,0.0004398825],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.4343551,0.0001754686,0.167163,0.001599498,0.0001970619,0.0004611191,0.000001959902,0.0001504309,0.3958964],"genre_scores_gemma":[0.9839355,0.000005345049,0.01280672,0.0003596995,0.00006655652,0.00003229168,9.60206e-7,0.000007291688,0.00278567],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.6791266,"threshold_uncertainty_score":0.8300546,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1519790996233589,"score_gpt":0.4051567908166698,"score_spread":0.2531776911933109,"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."}}