{"id":"W4399115506","doi":"10.48550/arxiv.2405.16325","title":"SLoPe: Double-Pruned Sparse Plus Lazy Low-Rank Adapter Pretraining of LLMs","year":2024,"lang":"en","type":"preprint","venue":"arXiv (Cornell University)","topic":"Network Packet Processing and Optimization","field":"Computer Science","cited_by":1,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Natural Sciences and Engineering Research Council of Canada; Department of Mechanical Engineering, University of Texas at Austin; Alliance de recherche numérique du Canada; DeepMind; Canada Research Chairs; University of Texas at Austin; National Science Foundation","keywords":"Adapter (computing); Rank (graph theory); Computer science; Psychology; Mathematics; Combinatorics; Computer hardware","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":[0.0007335134,0.002047299,0.001117103,0.0006099307,0.0006905323,0.001104888,0.003096246,0.00163512,0.01246849],"category_scores_gemma":[0.005816768,0.0009990742,0.0009924467,0.0005274857,0.0008269535,0.002141353,0.002309748,0.003749832,0.006116838],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0007098226,"about_ca_system_score_gemma":0.002393059,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.01115787,"about_ca_topic_score_gemma":0.03143081,"domain_scores_codex":[0.9993149,0.0001111865,0.00003619568,0.0001779505,0.000220495,0.0001392307],"domain_scores_gemma":[0.9986537,0.0004117154,0.00008757771,0.0004380431,0.0003099107,0.00009907098],"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":[0.0006989925,0.0004483206,0.003186503,0.0002797426,0.0001860436,0.0004171794,0.0002407476,0.3850046,0.02772416,0.01121878,0.04133526,0.5292597],"study_design_scores_gemma":[0.00003969649,0.00006448726,0.000236435,0.00001357668,0.00001430233,0.00004832374,0.00002393883,0.9878805,0.004748521,0.004420936,0.002495116,0.00001423847],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.02155031,0.0002863725,0.9516997,0.0004339683,0.0001943279,0.0001267772,0.0003855187,0.02172384,0.003599148],"genre_scores_gemma":[0.3656495,0.0002360742,0.6099486,0.001234865,0.0001876152,0.0004546923,0.003187911,0.002956392,0.01614443],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.01246849,"threshold_uncertainty_score":0.04171127,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.07694164391039529,"score_gpt":0.1976297395737091,"score_spread":0.1206880956633138,"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."}}