{"id":"W4416369164","doi":"10.48550/arxiv.2510.02613","title":"ElasticMoE: An Efficient Auto Scaling Method for Mixture-of-Experts Models","year":2025,"lang":"en","type":"preprint","venue":"ArXiv.org","topic":"Big Data and Digital Economy","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Huawei Technologies (Canada)","funders":"","keywords":"Scaling; Cloud computing; Inference; Provisioning; Throughput; Pipeline transport","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.002124082,0.001833203,0.001318142,0.001260783,0.0007598958,0.001553735,0.00280573,0.001603258,0.0100751],"category_scores_gemma":[0.00945472,0.001368767,0.002068759,0.0008858247,0.0007504434,0.00260656,0.002685864,0.003383407,0.004011764],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00107642,"about_ca_system_score_gemma":0.001595218,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00990314,"about_ca_topic_score_gemma":0.02020013,"domain_scores_codex":[0.9989362,0.0002600053,0.00006823127,0.0002904937,0.0003373111,0.0001076304],"domain_scores_gemma":[0.9978063,0.001263894,0.000123162,0.0003338408,0.0003622368,0.0001106602],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0003262284,0.0001615352,0.001987963,0.0002226533,0.0002609068,0.0002584472,0.0002007046,0.6832724,0.005357273,0.0128783,0.01512952,0.2799442],"study_design_scores_gemma":[0.00001085906,0.000008043635,0.00005700836,0.000005087436,0.000005951374,0.00001615172,0.00001106474,0.9938957,0.0005818136,0.004016418,0.001385594,0.000006278205],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.006969022,0.0003387234,0.979733,0.0001800457,0.00009549586,0.00007897655,0.0002747168,0.01092055,0.001409543],"genre_scores_gemma":[0.1928026,0.0003307693,0.7936938,0.000629525,0.0001583085,0.0004213011,0.001990821,0.004336872,0.005636131],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.0100751,"threshold_uncertainty_score":0.03370452,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.08906862300522915,"score_gpt":0.3320609641085206,"score_spread":0.2429923411032914,"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."}}