{"id":"W6979368922","doi":"","title":"Difficulty-Based Preference Data Selection by DPO Implicit Reward Gap","year":2025,"lang":"en","type":"article","venue":"ArXiv.org","topic":"Recommender Systems and Techniques","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Bundesministerium für Bildung und Forschung; Government of Canada; Canadian Institute for Advanced Research","keywords":"Preference; Selection (genetic algorithm); Preference learning; Reinforcement learning; Model selection; Revealed preference","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":[0.00688003,0.001843436,0.002508411,0.00188865,0.001016212,0.002324151,0.003591738,0.002032507,0.005320567],"category_scores_gemma":[0.03187697,0.0009866684,0.001614597,0.002526631,0.001395117,0.006197963,0.005548969,0.004098588,0.003165562],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001042121,"about_ca_system_score_gemma":0.001946954,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.001584889,"about_ca_topic_score_gemma":0.00298783,"domain_scores_codex":[0.9921877,0.004261841,0.0004861815,0.001504203,0.001224731,0.0003353303],"domain_scores_gemma":[0.9855471,0.008386481,0.0008350009,0.003419171,0.001241405,0.0005709372],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"theoretical_or_conceptual","study_design_scores_codex":[0.001340739,0.001250161,0.01216541,0.0007882701,0.0003865107,0.0004190008,0.0008957735,0.1681009,0.01609264,0.02867599,0.02245472,0.7474298],"study_design_scores_gemma":[0.000200467,0.0003018853,0.001154097,0.00004683609,0.00005484255,0.0002093271,0.0002133244,0.9462534,0.005183035,0.04123924,0.005079341,0.00006431604],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.04949458,0.0006498756,0.9412252,0.0006821308,0.000127718,0.0003435554,0.0007522551,0.004192844,0.002531773],"genre_scores_gemma":[0.5163457,0.0002292701,0.4738773,0.0008384158,0.0001897329,0.0007738809,0.003489136,0.0009350021,0.003321527],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.00688003,"threshold_uncertainty_score":0.03638554,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1030329214442541,"score_gpt":0.3095350622575327,"score_spread":0.2065021408132787,"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."}}