Reinforcement learning (RL) is a critical stage in post-training large language models (LLMs), involving repeated interaction between rollout generation, reward evaluation, and centralized learning. Distributing rollout execution offers opportunities to leverage more cost-efficient inference resources, but introduces challenges in wide-area coordination and policy dissemination. We present ECHO-2, a distributed RL framework for post-training with remote inference workers and non-negligible dissemination latency. ECHO-2 combines centralized learning with distributed rollouts and treats bounded policy staleness as a user-controlled parameter, enabling rollout generation, dissemination, and training to overlap. We introduce an overlap-based capacity model that relates training time, dissemination latency, and rollout throughput, yielding a practical provisioning rule for sustaining learner utilization. To mitigate dissemination bottlenecks and lower cost, ECHO-2 employs peer-assisted pipelined broadcast and cost-aware activation of heterogeneous workers. Experiments on GRPO post-training of LLMs ranging from 4B to 32B parameters under real wide-area bandwidth regimes show that ECHO-2 significantly improves cost efficiency while preserving RL reward comparable to strong baselines.
@article{arxiv.2602.02192,
title = {ECHO-2: A Large-Scale Distributed Rollout Framework for Cost-Efficient Reinforcement Learning},
author = {Jingwei Song and Meng Chen and Jie Xiao and Qingnan Ren and Jiaqi Huang and Yangshen Deng and Chris Tong and Wanyi Chen and Suli Wang and Zhisheng Chen and Ziqian Bi and Shuo Lu and Yiqun Duan and Xu Wang and Rymon Yu and Lynn Ai and Eric Yang and Tianyu Shi},
journal= {arXiv preprint arXiv:2602.02192},
year = {2026}
}