MoEs Are Stronger than You Think: Hyper-Parallel Inference Scaling with RoE

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Zibakhsh, Soheil, Samragh, Mohammad, Nishu, Kumari, Hannah, Lauren, Kundu, Arnav, Cho, Minsik
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866911208211218432
author Zibakhsh, Soheil
Samragh, Mohammad
Nishu, Kumari
Hannah, Lauren
Kundu, Arnav
Cho, Minsik
author_facet Zibakhsh, Soheil
Samragh, Mohammad
Nishu, Kumari
Hannah, Lauren
Kundu, Arnav
Cho, Minsik
contents The generation quality of large language models (LLMs) is often improved by utilizing inference-time sequence-level scaling methods (e.g., Chain-of-Thought). We introduce hyper-parallel scaling, a complementary framework that improves prediction quality at the token level. Hyper-parallel scaling computes and aggregates multiple output proposals for a single token from the model. We implement this concept in Mixture-of-Experts (MoE) models, which we refer to as Roster of Experts (RoE). RoE is a training-free inference algorithm that turns a single MoE into a dynamic ensemble of MoEs. RoE injects controlled stochasticity into the expert routing mechanism, enabling it to sample multiple diverse experts for each token and aggregate their outputs for a more accurate final prediction. To overcome the computational cost, we introduce an efficient batching strategy and a specialized KV-caching mechanism that minimizes compute and memory overhead. For example, RoE enables a 7B MoE model to match the performance of a 10.5B MoE model while using 30% less compute for inference. These gains are achieved without any fine-tuning of model parameters.
format Preprint
id arxiv_https___arxiv_org_abs_2509_17238
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle MoEs Are Stronger than You Think: Hyper-Parallel Inference Scaling with RoE
Zibakhsh, Soheil
Samragh, Mohammad
Nishu, Kumari
Hannah, Lauren
Kundu, Arnav
Cho, Minsik
Artificial Intelligence
Computation and Language
Machine Learning
The generation quality of large language models (LLMs) is often improved by utilizing inference-time sequence-level scaling methods (e.g., Chain-of-Thought). We introduce hyper-parallel scaling, a complementary framework that improves prediction quality at the token level. Hyper-parallel scaling computes and aggregates multiple output proposals for a single token from the model. We implement this concept in Mixture-of-Experts (MoE) models, which we refer to as Roster of Experts (RoE). RoE is a training-free inference algorithm that turns a single MoE into a dynamic ensemble of MoEs. RoE injects controlled stochasticity into the expert routing mechanism, enabling it to sample multiple diverse experts for each token and aggregate their outputs for a more accurate final prediction. To overcome the computational cost, we introduce an efficient batching strategy and a specialized KV-caching mechanism that minimizes compute and memory overhead. For example, RoE enables a 7B MoE model to match the performance of a 10.5B MoE model while using 30% less compute for inference. These gains are achieved without any fine-tuning of model parameters.
title MoEs Are Stronger than You Think: Hyper-Parallel Inference Scaling with RoE
topic Artificial Intelligence
Computation and Language
Machine Learning
url https://arxiv.org/abs/2509.17238