{"id":"W7103204945","doi":"","title":"What's In My Human Feedback? Learning Interpretable Descriptions of Preference Data","year":2025,"lang":"","type":"article","venue":"ArXiv.org","topic":"Explainable Artificial Intelligence (XAI)","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Canadian Institute for Advanced Research; Open Philanthropy Project; National Science Foundation","keywords":"Preference; Diversity (politics); Preference elicitation; Preference learning; Work (physics)","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":"codex-gemma-dda1882f352a","candidate_categories":["metaepi_narrow","open_science"],"consensus_categories":[],"category_scores_codex":[0.001873685,0.000495224,0.0007444617,0.0007585824,0.0006079671,0.0009915299,0.006118343,0.0002848665,0.0003490217],"category_scores_gemma":[0.001067396,0.0005748437,0.0001337754,0.00313224,0.000567366,0.008710913,0.005112764,0.00124769,0.0004290402],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.000295442,"about_ca_system_score_gemma":0.0006463185,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002031774,"about_ca_topic_score_gemma":0.001964744,"domain_scores_codex":[0.994331,0.0006268914,0.001634692,0.001801848,0.0004838534,0.001121731],"domain_scores_gemma":[0.9947648,0.0004147591,0.0005030219,0.003672617,0.0004625275,0.0001822814],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0001146695,0.001831519,0.7426101,0.0007182325,0.0002670782,0.00008879056,0.02537902,0.006217018,0.058574,0.05571937,0.002113156,0.106367],"study_design_scores_gemma":[0.001263428,0.001119657,0.3201611,0.01247191,0.0002678934,0.00001546289,0.02983164,0.4333792,0.1562046,0.01670374,0.02625817,0.002323129],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.8835304,0.002927811,0.1028485,0.001576662,0.002651181,0.0006772112,0.00001120034,0.0001351526,0.005641838],"genre_scores_gemma":[0.9896781,0.001056412,0.001257303,0.0002772163,0.00009247468,0.00004004338,0.00003249282,0.0000288256,0.007537163],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.4271622,"threshold_uncertainty_score":0.9996703,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1960479997698643,"score_gpt":0.3504283725767246,"score_spread":0.1543803728068603,"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."}}