{"id":"W7094976913","doi":"","title":"ALF : Un jeu de données d'analogies françaises à grain fin pour l'évaluation de la connaissance lexicale des grands modèles de langue","year":2025,"lang":"fr","type":"article","venue":"HAL (Le Centre pour la Communication Scientifique Directe)","topic":"Qualitative Comparative Analysis Research","field":"Social Sciences","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Université de Montréal","funders":"","keywords":"Context (archaeology); Focus (optics); Pragmatics; Denotation (semiotics)","routes":{"ca_aff":true,"ca_fund":false,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":["metaresearch","metaepi_narrow","sts","insufficient_payload"],"consensus_categories":["sts"],"category_scores_codex":[0.02287239,0.0003014961,0.0004475153,0.0003052523,0.001703717,0.0005097294,0.001245551,0.0002880924,0.0009575317],"category_scores_gemma":[0.0145699,0.0003266825,0.0002903931,0.001837516,0.004204178,0.0003633753,0.0003207326,0.0005291601,0.00008112349],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0004789566,"about_ca_system_score_gemma":0.001290668,"about_ca_topic_candidate":true,"about_ca_topic_consensus":true,"about_ca_topic_score_codex":0.01765998,"about_ca_topic_score_gemma":0.02073191,"domain_scores_codex":[0.9574513,0.0400303,0.0005218326,0.0005753561,0.000614629,0.0008066276],"domain_scores_gemma":[0.9838241,0.01266787,0.0002734737,0.0007572721,0.002211079,0.0002661726],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"theoretical_or_conceptual","study_design_scores_codex":[0.00003125372,0.0006154186,0.04060282,0.0001553803,0.0002439191,0.000005394056,0.1256361,0.0005428343,0.01070826,0.7485654,0.00973352,0.06315964],"study_design_scores_gemma":[0.00150782,0.000003419367,0.1918284,0.002611348,0.0002496596,0.000004413581,0.01924316,0.07782584,0.1272609,0.4871035,0.09162404,0.0007375611],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":"theoretical_or_conceptual","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.3409496,0.01460238,0.4471329,0.03972717,0.00007243545,0.0002892889,0.00005027412,0.0001292289,0.1570468],"genre_scores_gemma":[0.9158817,0.003572796,0.04033343,0.0002174473,0.00004290867,0.00007491702,0.00006096963,0.00002304206,0.03979278],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.5749321,"threshold_uncertainty_score":0.9999557,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.075285955257089,"score_gpt":0.3647547896709896,"score_spread":0.2894688344139006,"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."}}