ResearchAug 19, 2026 · 01:594 sources
RadixArk Launches Miles RL Framework to Open Frontier Model Post-Training
From huggingnews.com
RadixArk has launched Miles v0.1, an open source reinforcement learning framework for large language models and multimodal systems, designed to improve training efficiency on NVIDIA and AMD hardware. The framework is already in use at IBM, Nebius AI, Modal, and Decagon AI, and was developed over nine months with 72 contributors and 1,326 commits. Miles uses SGLang as its rollout engine and Megatron-LM and FSDP for training, and was validated on models including DeepSeek V4, Qwen 3.8, Kimi K3, and GLM 5.2.
Read the sourceSign in to suggest
Sources
+ 3 more
- SUPPORT@lmsysorg“72 contributors have landed 1,326 commits, 85 GPU E2E CI tests, battle-testing Miles on frontier open models like Kimi K3, DeepSeek V4, Qwen 3.8, GLM 5.2”x.com
- SUPPORT@arcee_ai“glad to see another strong contribution to the open post-training ecosystem”x.com
- SUPPORT@willccbb“miles v0.1 ships with first-class support for training with `verifiers` envs”x.com
#open-source#llm#rl#multimodalMarkdown
Content history (3)
- 2026-10-16 · Summary · vi · Replaced stored textCác nhà phát triển hiện có thể dùng framework mã nguồn mở Miles v0.1 để triển khai reinfo…Các nhà phát triển hiện có thể sử dụng khung nguồn mở Miles v0.1 để triển khai học tăng c…
- 2026-10-16 · Title · vi · Replaced stored textRadixArk ra mắt Miles RL framework hỗ trợ hậu huấn luyện các mô hình tiên tiếnRadixArk ra mắt khung Miles RL để mở rộng mô hình AI sau khi huấn luyện
- 2026-10-04 · Summary · vi · Wording fixCác nhà phát triển hiện có thể dùng framework mã nguồn mở Miles v0.1 để triển khai reinfo…Các nhà phát triển hiện có thể dùng framework mã nguồn mở Miles v0.1 để triển khai reinfo…
One ranked edition, every morning
The stories that matter today. Free, cancel anytime.
Telegram @aidr_today