Xiaomi's CodeMidas turns open-source code into 5,545 RL environments, boosting coding agents up to 17%
eliebakouch · x · 2026-09-21
Xiaomi MiMo's CodeMidas: Scaling Coding RL Environments from Source Code
Xiaomi's MiMo team presents CodeMidas, an agentic pipeline that converts implemented functionality in existing codebases into executable RL environments, using source code alone—no reliance on issues or commits.
Agentic pipeline highlights:
- Agents explore code to write behavioral specifications and build execution-grounded tests
- Tasks are validated via execution checks and repeated solution rollouts
- Robustness is tested by letting agents try to cheat; difficulty measured via pass@n
Results:
- 5,545 training tasks from 3,185 open-source repos, spanning 23 languages and 15 domains
- Training MiMo-V2.5 with GRPO improves all five benchmarks: DeepSWE +11.7%, ProgramBench +17%, Terminal-Bench v2.1 +8.5%
- Ablations show more high-quality tasks is the key scaling axis
More from coding & agent
- Photoroom reclaims ~1,000 idle GPU hours a week with SkyPilot across Slurm and Kubernetes — skypilot_org · 2026-11-12
- Google launches Gemini Decisions API: scores 1,000 questions per pass, output tokens free — dean_rie · 2026-11-12
- Zed v1.28 adds background auto-fetch for current and default branch remotes — dean_rie · 2026-11-12
- Indie Britpop AI music video built with Nano Banana Pro, Kling, and free lip-sync via WAN2GP + LTX 2.5 — cocktailpeanut · 2026-11-12
- Cloudflare Talk: Swapping Some Agent LLM Calls for Fast Models Like Jev Gives Big Speedups — michellechen · 2026-11-12
- Engineer Curates 15 AI Agent Design Patterns Across Single-Agent, Multi-Agent and Feedback Tiers — MaryamMiradi · 2026-11-12