{"id":"W7161165964","doi":"10.1145/3805689.3812346","title":"Prompting from the bench: Large-scale pretraining is not sufficient to prepare LLMs for ordinary meaning analysis","year":2025,"lang":"","type":"article","venue":"ArXiv.org","topic":"Artificial Intelligence in Law","field":"Social Sciences","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Alliance de recherche numérique du Canada; Georgetown University; National Science Foundation","keywords":"Interpretation (philosophy); Argument (complex analysis); Scholarship; Meaning (existential); Leverage (statistics); Criminal trial","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":[0.005541363,0.001430809,0.0007552242,0.0005615363,0.0009060105,0.002955194,0.002156744,0.00199218,0.01200093],"category_scores_gemma":[0.03582031,0.0008305261,0.001061264,0.0006884785,0.001456142,0.006705439,0.002214639,0.00629243,0.006914792],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001822663,"about_ca_system_score_gemma":0.002350048,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.006351305,"about_ca_topic_score_gemma":0.01569398,"domain_scores_codex":[0.9967714,0.0016357,0.0001566585,0.0009157719,0.0003220681,0.0001984488],"domain_scores_gemma":[0.9765342,0.01692087,0.0007818641,0.003984312,0.001170146,0.0006085289],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.001477996,0.0009391958,0.02506905,0.0009039168,0.0003656773,0.0003982615,0.002084941,0.05211419,0.01403575,0.01050727,0.1365569,0.7555469],"study_design_scores_gemma":[0.0003282182,0.000748241,0.01809148,0.0007448187,0.0002088026,0.0003108217,0.003009694,0.7869642,0.02023113,0.1150805,0.0541025,0.0001795724],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.5392823,0.003201162,0.3309683,0.02385107,0.002665142,0.0009864689,0.007733669,0.0293045,0.06200744],"genre_scores_gemma":[0.878224,0.0003207116,0.09814462,0.00432469,0.0001669385,0.0004328965,0.0110639,0.0006910961,0.006631204],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.01200093,"threshold_uncertainty_score":0.04014707,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.09621491876583217,"score_gpt":0.3808325565161599,"score_spread":0.2846176377503278,"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."}}