Loading
Life sciences · Preprint
arXiv · September 8, 2026
Early or partial results. Treat as a signal, not a conclusion.
Miles v0.1 is an open-source software framework for large-scale reinforcement learning post-training, described as production-ready and supporting multiple training backends and RL paradigms. The report includes a single end-to-end case study on terminal-use coding tasks using a 744B parameter model, but provides no comparative validation, benchmark results, or evidence of efficacy relative to alternatives.
Preprint.
System supports multiple RL training modes: full-parameter RL, LoRA RL, on-policy distillation, supervised fine-tuning, and on-policy rollout-training alignment Case study: agentic RL on GLM-5.2 744B-A40B model running on 64 NVIDIA GB300 GPUs with median step time of 263 seconds over first 30 measured steps Rollout engine built on SGLang; trainer offers NVIDIA Megatron-LM and PyTorch FSDP backends; three weight-synchronization transports for different deployment topologies
Safety was not reported in the material analysed. Check the source before drawing any conclusion about harm.
The source did not state who this applies to in practice.
This is a technical systems report describing a software framework for reinforcement learning post-training, with a single case study but no comparative evaluation, experimental validation, or peer review.
Quoted from the source exactly as published.
Graded across the dimensions that decide whether you should act, each from what the source actually supports. There is no single score, and where a dimension was not assessed it says so.
We 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.
Taken from the source record, never inferred. Follow any of these and new work involving them reaches your briefing.