Reinforcement learning (RL) has been a critical paradigm for improving large language model (LLM) reasoning and agentic capabilities, but its efficiency remains a major bottleneck. Existing RL systems commonly rely on asynchronous execution to overlap rollout generation with policy optimization for enhanced efficiency. However, this approach faces inherent limitations: mild policy staleness offers only limited training efficiency gains, whereas aggressive asynchrony can degrade model quality and yields diminishing efficiency gains. The dynamic nature of LLM RL and multi-turn agentic workloads further challenges RL system design, as resource demands shift over time, making static resource allocation and generation strategy inefficient. This paper presents ReFlow, a system-algorithm co-designed RL framework for efficient LLM post-training. Its key insight is that, unlike existing RL frameworks which consume each trajectory only once, experience replay of recent trajectories enables effective sample reuse, thereby amortizing the high cost of rollout generation without incurring excessive policy staleness. To fully exploit this opportunity, ReFlow adopts a decoupled and adaptive architecture with three key designs: a dual replay data cycle supports efficient experience reuse while balancing training throughput and optimization stability; an elastic resource pool enables dynamic resource placement to accommodate evolving RL workloads; an adaptive scheduler coordinates generation strategies and resource allocation to improve resource utilization and boost generation throughput. Experiments across representative single-turn reasoning and multi-turn agentic RL tasks show that ReFlow improves training throughput by up to 2.06× over state-of-the-art synchronous and asynchronous RL systems, while maintaining comparable or better model performance.
Figures & tables
Figure 1: Typical LLM RL workflow: (a) synchronous RL; (b) asynchronous RL. W denotes different versions of weights.
Figure 2: GPU idle ratio (idle time / step time) under different resource allocations for Qwen3-4B training on 8 GPUs.
Figure 3: Training model performance under different staleness settings when training Qwen3-1.7B on 8 GPUs, where staleness denotes the number of training iterations between two consecutive parameter synchronizations.
Figure 4: Diverse and evolving generation characteristics under different training configurations, demonstrating the highly dynamic workload pattern of LLM RL.
Figure 5: Qwen3-4B on DAPO-MATH-17K. Naive experience replay slows convergence and degrades performance.
Figure 6: System overview of ReFlow.
Figure 7: Dataflow of ReFlow.
Symbol
Description
σ
Maximum fraction of replay data in a training batch.
τ
Loss weight controlling the relative contribution of replayed trajectories.
δ
Maximum data policy lag allowed for training.
Table 1: Hyperparameters in dual replay data cycle.
Figure 8: Cold-start overhead. (a) Total overhead under Qwen3 models. (b) Breakdown for Qwen3-32B on 128 GPUs.
Figure 9: Workflow of ReFlow’s resource rebalance. We use TP=2 for the rollout worker and FSDP group size=4 for the trainer as an example.
Reinforcement learning post-training adapts pretrained models with reward or feedback signals to improve reasoning, alignment, and interactive behavior.
Reinforcement learning (RL) has become a pivotal technology in the post-training phase of large language models (LLMs). Traditional task-collocated RL frameworks suffer from significant scalability bottlenecks, while task-separated RL frameworks face challenges in managing complex dataflows and resolving resource idling. Furthermore, most existing frameworks are tightly coupled with LLM training or inference engines, making them difficult to support custom-designed engines. To address these challenges, we propose AsyncFlow, an asynchronous streaming RL framework tailored for efficient post-training. Specifically, we introduce a distributed data storage and transfer module that provides panoramic data management and fine-grained scheduling capabilities in a fully streamed manner. This architecture inherently enables automated pipeline overlapping among RL tasks and dynamic load-balancing. Moreover, we propose an asynchronous producer-consumer workflow, which is engineered to minimize computational idleness by strategically deferring the parameter update process within staleness thresholds. Finally, the core capabilities of AsyncFlow are architecturally decoupled from underlying training and inference engines and encapsulated by service-oriented user interfaces, offering a modular and customizable user experience. Extensive experiments demonstrate an average throughput of 1.59x compared to the state-of-the-art baseline. The architecture presented in this work provides actionable insights for designing next-generation RL training systems.
Reinforcement learning (RL) has emerged as a standard post-training paradigm for shaping large language models (LLMs) into capable agents. In agentic RL, the rollout stage generates trajectories while invoking tools, producing long-tailed and non-stationary workloads that expose two fundamental challenges. First, due to the long-tailed response distribution, a small fraction of trajectories dominates rollout makespan.Second, rollout and training differ in their compute patterns, memory demands, and sensitivity to sequence length. As the policy evolves, shifts in the workload distribution further change their relative resource demands, making it difficult to maintain balanced execution across the two stages. We present Libra, an adaptive runtime for agentic RL post-training with two complementary components: (1) intra-stage scheduling via a Causality-Guided Bucket Scheduler that routes requests across execution buckets with different parallelism configurations, reducing delays from rollout stragglers; and (2) cross-stage coordination that dynamically reallocates workers between rollout and training as the workload changes. It moves workers between the two stages through a non-blocking protocol without interrupting ongoing training. Evaluated on a 48x NVIDIA A800 GPU cluster and a 160x Ascend 910B3 NPU cluster across three agentic benchmarks, Libra achieves up to 4.2x higher throughput and up to 2.7x faster reward convergence
Kaiwen Chen, Xin Tan, Jingzong Li +6
The Chinese University of Hong Kong · The Hang Seng University of Hong Kong · Independent Researcher
Reinforcement learning (RL) has become a critical paradigm for LLM post-training, yet the rollout phase -- accounting for 50--80% of total step time -- is bottlenecked by skewed generation: long-tailed trajectories indispensable for model performance block the entire training pipeline. Asynchronous training offers a natural remedy by overlapping generation with training, but introduces a fundamental tension between efficiency and algorithmic correctness. We identify three constraints in asynchronous training to preserve convergence: intra-trajectory policy consistency, data integrity, and bounded staleness. Existing approaches fail to intrinsically address the long-tailed trajectory problem, which is further exacerbated by the imbalance characteristic of Mix-of-Experts models, or deviate from the standard RL training formulation, thereby hindering model convergence. Therefore, we propose DORA (Dynamic ORchestration for Asynchronous Rollout), which addresses this challenge through algorithm-system co-design. DORA introduces multi-version streaming rollout, a novel asynchronous paradigm that maintains multiple policy versions concurrently -- simultaneously achieving full bubble elimination without compromising algorithmic constraints. Experimental results demonstrate that our DORA system achieves substantial improvements in throughput -- up to 2--3 times higher than state-of-the-art systems on open-source benchmarks -- without compromising convergence. Furthermore, in large-scale industrial applications with tens of thousands of accelerators, DORA accelerates RL training by 2--4 times compared to synchronous training across various scenarios. The resultant open-source models, LongCat-Flash-Thinking, exhibit competitive performance on complex reasoning benchmarks, matching the capability of most advanced LLMs.