{"id":"W7107999839","doi":"","title":"ViLU: Apprentissage d'incertitude langage et visuel pour la prédiction d'erreur","year":2025,"lang":"en","type":"article","venue":"HAL (Le Centre pour la Communication Scientifique Directe)","topic":"Artificial Intelligence in Education","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"École de Technologie Supérieure","funders":"Grand Équipement National De Calcul Intensif; Agence Nationale de la Recherche","keywords":"Reading (process); Reflection (computer programming); Focus (optics); Context (archaeology)","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":[],"consensus_categories":[],"category_scores_codex":[0.004602146,0.0002240408,0.0002017902,0.0002600484,0.0004359776,0.0006440365,0.001666013,0.0001333174,0.00007378004],"category_scores_gemma":[0.001520678,0.0002425482,0.0001198864,0.0009182876,0.0001545525,0.0006559521,0.0006180247,0.0003921664,0.0002081441],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001369407,"about_ca_system_score_gemma":0.0003505301,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0006006441,"about_ca_topic_score_gemma":0.001397225,"domain_scores_codex":[0.9941227,0.004034189,0.0004584582,0.0006717362,0.0003701381,0.0003428048],"domain_scores_gemma":[0.9952505,0.001720324,0.0002460016,0.001643267,0.0010142,0.00012577],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.000006601992,0.0009067355,0.00293859,0.00005103967,0.00004497894,0.000005421417,0.01099392,0.000121386,0.01496944,0.7445842,0.01026111,0.2151165],"study_design_scores_gemma":[0.0006895015,0.00000198748,0.04331483,0.001790167,0.00006814882,0.00003778818,0.0008639096,0.222402,0.4361474,0.06445651,0.229174,0.001053779],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.03336613,0.000242261,0.842434,0.02192232,0.0005695272,0.0002701318,0.000005366495,0.0004618795,0.1007284],"genre_scores_gemma":[0.9175644,0.0001470517,0.06643709,0.0003521827,0.00002282572,0.00006984836,0.00004893158,0.00002154056,0.01533613],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.8841983,"threshold_uncertainty_score":0.9890826,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01643879059458228,"score_gpt":0.2773303489592637,"score_spread":0.2608915583646814,"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."}}