cs.LGNov 13, 2026

RLRT^{\mathrm{RT}}: On-Device Reinforcement Learning Runtime

Authors: Zexin Li, Xinwei Gao, Arvind Easwaran

Organizations: Nanyang Technological University, Singapore

Abstract

On-device reinforcement learning runs a periodic actor and a learner on one embedded GPU shared with other workloads. Because the learner produces the actor's weights, existing designs stall the actor or let its policy drift. We present RLRT^{\mathrm{RT}}, an on-device runtime that makes this coupling a tunable parameter: a refresh bound limits actor releases between refreshes, and a GPU permit with slack-aware admission prioritises the actor and fills spare time with training atop any base scheduler. Under the real refresh cost on two Jetson boards, RLRT^{\mathrm{RT}} keeps a 16.8 M-parameter policy's deadline misses below 1%, against 85 to 100% without it.

Figures & tables

Explore similar work

Aug 11, 2026

TideRL: Boosting Agentic RL Goodput with Readiness-Aware Scheduling

Reinforcement learning (RL) for large language models is moving toward multi-turn agentic workloads, where rollout tasks repeatedly pause for external environments, resume with growing contexts, and finish at highly variable times. In this setting, RL training goodput, measured by training throughput, matters more than raw GPU occupancy: GPU waiting and repeated prefill recomputation are pure overhead. We present TideRL, a readiness-aware elastic RL system with Continuous Task Batching, Resource-Aware Ref-Actor Pipelining, and Elastic Resource Scaling. CTB preserves useful rollout state, RA2P\textrm{RA}^2\textrm{P} selects between decoupled streaming and colocated aggregation from the ready backlog and arrival interval, and ERS moves ranks between rollout and training using the same readiness signals. Across text-only and multi-modal agentic workloads, TideRL improves RL training goodput by up to 5.6×\times over synchronous baselines and over 33% over asynchronous baselines, while reaching similar task performance. It also improves KV cache hit rate by 1.58×\times, reduces per-step training time by up to 44.3%, and cuts total waiting time by up to 77.6%.
Jun 24, 2026

Finding the Time to Think: Learning Planning Budgets in Real-Time RL

Deliberating takes time. In real-time settings, that time is not free. Standard reinforcement learning (RL) sidesteps this as the environment waits indefinitely for the agent's decision. Instead, we study real-time RL environments where the environment progresses while waiting for the agent's action. Building on prior real-time formalizations, we introduce variable-delay real-time RL, where the agent chooses how long to deliberate at each decision point since the environment progresses. For the planning agents we use, the right delay is state-dependent, and naively planning how long to plan can paralyze the agent. We instead approach this setting by training a lightweight gating policy on top of a planner to select state-dependent planning budgets. Across real-time Pac-Man, Tetris, Snake, Speed Hex, and Speed Go, our gating policy outperforms fixed-budget and heuristic baselines, and transfers to a real-time setup where the environment and agent run on two different GPUs.
Apr 10, 2026

TensorHub: Scalable and Elastic Weight Transfer for LLM RL Training

Modern LLM reinforcement learning (RL) workloads require a high-performance weight transfer system to scale training across heterogeneous compute resources. However, efficiently transferring terabyte-scale model weights across thousands of GPUs remains challenging because the system must accommodate clusters that dynamically scale up and down while keeping coordination, data movement, and storage overhead low. We introduce Reference-Oriented Storage (ROS), a new storage abstraction for RL weight transfer that exploits highly replicated model weights in place. ROS presents the illusion that certain versions of the model weights are stored and can be fetched on demand. Underneath, ROS does not physically store any copies of the weights; instead, it tracks the workers that hold these weights on GPUs for inference. Upon request, ROS directly uses them to serve reads. We build TensorHub, a production-quality system that instantiates the ROS idea with topology-aware transfer, model-parallel consistency, and fault tolerance. Evaluation shows that TensorHub saturates RDMA bandwidth and adapts to three distinct rollout workloads with minimal engineering effort. Specifically, TensorHub reduces total GPU stall time by up to 6.7x for standalone rollouts, accelerates weight updates for elastic rollouts by up to 4.8x, and cuts cross-datacenter rollout stall time by up to 19x. TensorHub has been deployed in ByteDance production to support cutting-edge RL training.