titleMiles v0.1: Production-Level Post-Training
authorsRadixArk, Tom Chen, Mao Cheng, Shi Dong, Kangrui Du, Yanbin Jiang, Jiajun Li, Yiming Li, Tao Lin, Yusheng Su, Andy Ye, Yueming Yuan, Zhichen Zeng
date2026-09-08
venuearXiv (preprint, not peer reviewed)
idarXiv:2609.08368
affiliationRadixArk
upvotes47 (Hugging Face Daily Papers, 2026-11-01)
linksabs pdf hf code
abstractWe present Miles v0.1, a full-stack, production-ready system for frontier post-training. Building upon the clean design of slime, Miles designs each stage of the reinforcement-learning (RL) training loop around a single principle: components should be verified, clean, and customizable. With accuracy, efficiency, reliability, and scalability as first-class goals, Miles aims to make frontier-scale RL accessible to researchers and enterprises alike. This report walks through the system end to end: rollout engines built on SGLang, a trainer with a choice of two backends (NVIDIA Megatron-LM and PyTorch FSDP), and three weight-synchronization transports for different deployment topologies. Beyond full-parameter RL, Miles also supports LoRA RL, on-policy distillation, supervised fine-tuning, and true-on-policy rollout-training alignment, and extends the same architecture to diffusion models. We close with an end-to-end case study: fully asynchronous agentic RL on a GLM-5.2 744B-A40B model over terminal-use coding tasks, running on 64 NVIDIA GB300 GPUs with a median step time of 263 seconds over the first 30 measured steps. Miles is open-sourced at https://github.com/radixark/miles, with the project website at https://miles.radixark.com.

cite

@misc{radixark2026miles,
  title={Miles v0.1: Production-Level Post-Training},
  author={RadixArk and Tom Chen and Mao Cheng and Shi Dong and Kangrui Du and Yanbin Jiang and Jiajun Li and Yiming Li and Tao Lin and Yusheng Su and Andy Ye and Yueming Yuan and Zhichen Zeng},
  year={2026},
  eprint={2609.08368},
  archivePrefix={arXiv},
  url={https://arxiv.org/abs/2609.08368}
}