{"id":"W4416078514","doi":"10.1145/3746252.3760808","title":"Approximating Gradient-Based Influence for Scalable Instruction Data Selection","year":2025,"lang":"","type":"article","venue":"","topic":"Natural Language Processing Techniques","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Toronto; York University","funders":"Natural Sciences and Engineering Research Council of Canada","keywords":"Scalability; Selection (genetic algorithm); Fraction (chemistry); Regression; Sample (material); Model selection; Computation; Feature selection","routes":{"ca_aff":true,"ca_fund":true,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.001144356,0.001419753,0.001300015,0.001237587,0.0005394393,0.001128325,0.001499336,0.0007939009,0.003560284],"category_scores_gemma":[0.0101108,0.0005603558,0.0009340793,0.001063058,0.0006995368,0.00183413,0.001340221,0.001609851,0.002590812],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0009439555,"about_ca_system_score_gemma":0.001594978,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.007766535,"about_ca_topic_score_gemma":0.01524187,"domain_scores_codex":[0.998942,0.000227947,0.00006545374,0.0003061949,0.0003376563,0.0001207835],"domain_scores_gemma":[0.9979544,0.001092629,0.0001152416,0.0003260954,0.0004219417,0.00008968641],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0005936003,0.0002477942,0.005663611,0.0002937361,0.0001067999,0.0001664686,0.0002610871,0.2554121,0.02646954,0.00654182,0.01785165,0.6863918],"study_design_scores_gemma":[0.00002378951,0.00005266656,0.0005650709,0.000008338489,0.00001126073,0.00003177086,0.00002345556,0.9865861,0.0070045,0.00357725,0.002102721,0.00001309371],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.05222367,0.0009205783,0.9214607,0.0003039931,0.0001349329,0.0001726885,0.0006419946,0.02128152,0.00285991],"genre_scores_gemma":[0.5442448,0.0003261732,0.4421386,0.0004411124,0.0001555165,0.0004643973,0.003660682,0.003011985,0.00555677],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.007766535,"threshold_uncertainty_score":0.01544267,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02190326696587988,"score_gpt":0.3060999106236422,"score_spread":0.2841966436577623,"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."}}