{"id":"W6929364148","doi":"10.48448/rqy4-hv57","title":"BAPO: Base-Anchored Preference Optimization for Personalized Alignment in LLMs","year":2024,"lang":"en","type":"other","venue":"Underline Science Inc.","topic":"Epigenetics and DNA Methylation","field":"Biochemistry, Genetics and Molecular Biology","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Kootenay Association for Science & Technology","funders":"","keywords":"Preference; Forgetting; Preference elicitation; Constraint (computer-aided design); Preference learning; Budget constraint","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":[0.000477211,0.0002030475,0.0001708318,0.0002641114,0.00004401078,0.00007540847,0.0002946378,0.0002303694,0.0001775317],"category_scores_gemma":[0.0001112557,0.000188186,0.00005690976,0.0002661094,0.0002330899,0.000003872203,0.0001058693,0.00007840686,0.00001656895],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0000574809,"about_ca_system_score_gemma":0.0002508498,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00005003633,"about_ca_topic_score_gemma":0.0001953633,"domain_scores_codex":[0.998485,0.0000276063,0.0002265611,0.0007013049,0.0002750239,0.0002844566],"domain_scores_gemma":[0.9993921,0.000009563395,0.0001153543,0.0003246418,0.00007908961,0.0000792725],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"bench_or_experimental","study_design_gemma":"not_applicable","study_design_scores_codex":[0.0002107923,0.0006345765,0.0006555351,0.0009800377,0.0002183821,0.00001150365,0.00053678,0.03733047,0.75173,0.009782822,0.1601172,0.03779188],"study_design_scores_gemma":[0.00166741,0.0006500018,0.00003964987,0.0003726086,0.00008810387,0.000002065401,0.0001672354,0.1062989,0.08250051,0.002024036,0.8051667,0.001022788],"study_design_candidate":"not_applicable","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"other","genre_scores_codex":[0.004944914,0.02998513,0.7182789,0.0009793431,0.003284728,0.004338064,0.000758394,0.000194083,0.2372364],"genre_scores_gemma":[0.1302914,0.004041531,0.2010394,0.000451148,0.001653278,0.0005010106,0.002799271,0.0008024646,0.6584205],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.6692295,"threshold_uncertainty_score":0.7674001,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03456480475705707,"score_gpt":0.3046772097784516,"score_spread":0.2701124050213945,"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."}}