{"id":5922,"date":"2026-09-27T22:08:58","date_gmt":"2026-09-27T13:08:58","guid":{"rendered":"https:\/\/localmodelwatch.tsuchitsuchi.com\/2026\/09\/27\/xiaomimimo-agentic-rl-training-environment\/"},"modified":"2026-09-28T02:14:26","modified_gmt":"2026-09-27T17:14:26","slug":"xiaomimimo-agentic-rl-training-environment","status":"publish","type":"post","link":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/09\/27\/xiaomimimo-agentic-rl-training-environment\/","title":{"rendered":"XiaomiMiMo Releases Agentic RL Training Environment and Dataset"},"content":{"rendered":"

<\/p>\n

At a Glance<\/h2>\n
\n\n\n\n\n\n\n\n
Item<\/th>\nValue<\/th>\n<\/tr>\n<\/thead>\n
Repository<\/td>\nXiaomiMiMo\/MiMo-V2.6-RL-oss<\/a><\/td>\n<\/tr>\n
Published<\/td>\n2026-09-25<\/td>\n<\/tr>\n
Source type<\/td>\nPrimary source (the publisher itself)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n

Values determined by this site’s code at collection time. Dates are JST.<\/em><\/p>\n

<\/p>\n

Overview<\/h2>\n

XiaomiMiMo has released the dataset “XiaomiMiMo\/MiMo-V2.6-RL-oss” for building reinforcement learning training environments for LLM agents (Agentic RL), along with a fork repository of the reinforcement learning framework verl named “XiaomiMiMo\/verl”. The target model specified is “MiMo-V2.6-Distill-Qwen-9B”, and reproduction code for reinforcement learning environments across five domains (Code, Cyber, General, Visual, and Music) based on verl (0.9.0.dev) is presented.<\/p>\n

This material outlines the specific environmental setup and implementation design of verifiers to reproduce the detailed training recipe described in Section 7 of the technical report “MiMo-V2.6: Scaling Reinforcement Learning Towards Self-Improvement” regarding scaling reinforcement learning methods for self-improvement.<\/p>\n

Claims and Evidence<\/h2>\n

According to the presenter, the reproduction infrastructure for agentic reinforcement learning is structured as five types of RL environments combining domain-specific verification mechanisms ( verifiers) and task families. The correspondence of each environment provided within the fork repository of verl (0.9.0.dev) is as follows:<\/p>\n

\n

\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\n\n\n\n\n\n\n
Domain<\/th>\nTask Family<\/th>\nVerifier<\/th>\nLaunch script<\/th>\nSettings<\/th>\n<\/tr>\n<\/thead>\n
Code<\/td>\nSoftware engineering<\/td>\nExecutable tests<\/td>\nscripts\/code\/train.sh<\/code><\/td>\nscripts\/code\/env.example<\/code><\/td>\n<\/tr>\n
Cyber<\/td>\nVulnerability reproduction<\/td>\nRule checks<\/td>\nscripts\/arvo\/arvo.sh<\/code><\/td>\nscripts\/arvo\/arvo.env.example<\/code><\/td>\n<\/tr>\n
General<\/td>\nKnowledge work<\/td>\nRubric-based judging<\/td>\nscripts\/general\/general.sh<\/code><\/td>\nscripts\/general\/general.env.example<\/code><\/td>\n<\/tr>\n
Visual<\/td>\nWeb development<\/td>\nVisual grading<\/td>\nscripts\/design\/webdev.sh<\/code><\/td>\nscripts\/design\/webdev.env.example<\/code><\/td>\n<\/tr>\n
Music<\/td>\nSymbolic music composition<\/td>\nRule checks<\/td>\nscripts\/design\/music.sh<\/code><\/td>\nscripts\/design\/music.env.example<\/code><\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n

The characteristics of the task families and verification mechanisms in each domain are organized as follows:<\/p>\n