{"id":"W7103204388","doi":"","title":"Metis-SPECS: Decoupling Multimodal Learning via Self-distilled Preference-based Cold Start","year":2025,"lang":"","type":"article","venue":"ArXiv.org","topic":"Multimodal Machine Learning Applications","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"","keywords":"Generalization; Reinforcement learning; Task (project management); Task analysis; Verifiable secret sharing; Memorization","routes":{"ca_aff":false,"ca_fund":false,"ca_venue":false,"about_ca":true,"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":[0.002403985,0.002321634,0.001823288,0.0007151002,0.0007774353,0.001408256,0.004665705,0.001992592,0.005350088],"category_scores_gemma":[0.007726009,0.001086602,0.001499871,0.0007577231,0.001749934,0.003883796,0.00497183,0.004812116,0.002080173],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001331044,"about_ca_system_score_gemma":0.002195137,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.004306311,"about_ca_topic_score_gemma":0.007648588,"domain_scores_codex":[0.998461,0.0004722775,0.00008468575,0.0005228169,0.0002801569,0.0001790841],"domain_scores_gemma":[0.9971432,0.001313912,0.0001675885,0.0007052966,0.0004515155,0.0002184794],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0008229799,0.0005390287,0.00376494,0.000437798,0.0002327866,0.0002534814,0.0004317799,0.4924994,0.01598595,0.01973126,0.01165871,0.4536419],"study_design_scores_gemma":[0.00003992312,0.0001518613,0.0001797518,0.00001962546,0.00001810812,0.00003204644,0.00002503278,0.981174,0.003814795,0.0133724,0.001152099,0.00002041293],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.03743405,0.0005396732,0.9487197,0.0004095873,0.0001075899,0.0001694666,0.0003304632,0.008705429,0.003584123],"genre_scores_gemma":[0.6671271,0.0002496425,0.3186533,0.001404233,0.00009508526,0.0006368184,0.001966495,0.001191682,0.00867568],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.005350088,"threshold_uncertainty_score":0.01789778,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03630322338093774,"score_gpt":0.2811754722599381,"score_spread":0.2448722488790003,"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."}}