Miles v0.1: Production-Level Post-Training

Thu, 10 Sep 2026 00:00:00 +0900

Paper

Miles v0.1: A “verified, clean, and scalable” full-stack system for frontier RL post-training

TL;DR: Miles is a full-stack RL post-training framework that resolves the train-rollout mismatch at the system level — the mismatch that arises when rollout generation and training run on different engines, kernels, and precisions. It unifies SGLang-based rollout, a Megatron-LM/FSDP trainer, and three weight-synchronization transports (broadcast/P2P/disk-delta), while also guaranteeing token exactness (TITO) and expert-routing replay (R3). As an end-to-end case study, it trains GLM-5.2 744B-A40B on 64 GB300 GPUs with fully asynchronous agentic RL, achieving a median step time of 263 seconds (source: §9.2, Fig. 5).

Training-Efficiency on Jaehun's Blog

Miles v0.1: Production-Level Post-Training

Miles v0.1: A “verified, clean, and scalable” full-stack system for frontier RL post-training