{"id":"arxiv/2610.21363","url":"https://arxiv.org/abs/2610.21363","tags":["math.ST","cs.LG","stat.ML"],"score":0,"title":"Two Sample Sizes in Test-Time LLM Ensembling: Buffered ERM and Minimax Theory","domain":"math.ST","source":"arxiv","authors":["Zhi Zhang","Arash A. Amini"],"reviews":0,"subject":"math.ST","summary":"Test-time scaling can improve the performance of large language models by generating multiple answers to a question and aggregating them. Weighted ensembles extend this idea by drawing answers from several models, with the weights determining how often each model is used. Learning these weights from a training dataset involves two sample sizes: $Q$ labeled questions and $N$ generations per model and question. We study their interplay through the clean risk, the population error of the ensemble's infinite-generation voting rule. Under a margin condition with exponent $β$, we show that…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-10-15","reactions":{}}