Salvato in:
Dettagli Bibliografici
Autori principali: Yu, Timothy Tin Long, Singh, Gursimran, Shi, Ge, Sadri, Hanieh, Zhang, Yong, Fan, Zhenan
Natura: Preprint
Pubblicazione: 2026
Soggetti:
Accesso online:https://arxiv.org/abs/2605.08527
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866911664656351232
author Yu, Timothy Tin Long
Singh, Gursimran
Shi, Ge
Sadri, Hanieh
Zhang, Yong
Fan, Zhenan
author_facet Yu, Timothy Tin Long
Singh, Gursimran
Shi, Ge
Sadri, Hanieh
Zhang, Yong
Fan, Zhenan
contents Reinforcement Learning from Verifiable Rewards (RLVR) has significantly improved the reasoning capabilities of large language models (LLMs), particularly in multi-turn agentic settings involving environment interaction like tool use. However, fine-tuning such models remains prohibitively expensive due to high computational requirements, limiting accessibility. We propose MARLaaS (Multi-tenant Asynchronous RL as a Service), a system for concurrent RL fine-tuning across multiple users and tasks. Our approach is based on two key ideas: (1) sharing a base model across tenants using lightweight LoRA adapters, and (2) a disaggregated asynchronous architecture that decouples rollout generation, environment interaction, and policy training into independently scheduled stages. This design enables tasks to progress through the RL pipeline at their own pace in an event-driven manner, reducing cross-task interference, idle time, and end-to-end latency. In multi-task settings (we report up to 32 concurrent tasks), MARLaaS achieves single-task state-of-the-art performance while improving accelerator utilization by up to 4.3x and reducing end-to-end training time by 85%.
format Preprint
id arxiv_https___arxiv_org_abs_2605_08527
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle MARLaaS: Multi-Tenant Asynchronous Reinforcement Learning as a Service
Yu, Timothy Tin Long
Singh, Gursimran
Shi, Ge
Sadri, Hanieh
Zhang, Yong
Fan, Zhenan
Distributed, Parallel, and Cluster Computing
Artificial Intelligence
Reinforcement Learning from Verifiable Rewards (RLVR) has significantly improved the reasoning capabilities of large language models (LLMs), particularly in multi-turn agentic settings involving environment interaction like tool use. However, fine-tuning such models remains prohibitively expensive due to high computational requirements, limiting accessibility. We propose MARLaaS (Multi-tenant Asynchronous RL as a Service), a system for concurrent RL fine-tuning across multiple users and tasks. Our approach is based on two key ideas: (1) sharing a base model across tenants using lightweight LoRA adapters, and (2) a disaggregated asynchronous architecture that decouples rollout generation, environment interaction, and policy training into independently scheduled stages. This design enables tasks to progress through the RL pipeline at their own pace in an event-driven manner, reducing cross-task interference, idle time, and end-to-end latency. In multi-task settings (we report up to 32 concurrent tasks), MARLaaS achieves single-task state-of-the-art performance while improving accelerator utilization by up to 4.3x and reducing end-to-end training time by 85%.
title MARLaaS: Multi-Tenant Asynchronous Reinforcement Learning as a Service
topic Distributed, Parallel, and Cluster Computing
Artificial Intelligence
url https://arxiv.org/abs/2605.08527