Xiaomi's CodeMidas turns source code into RL environments: 5,545 tasks, 23 languages
teortaxesTex · x · 2026-09-21
A paper on Xiaomi's CodeMidas: rather than relying on issues or commits, it uses agents to extract executable tasks directly from open-source codebases to build RL training environments — generating 5,545 tasks across 23 languages.
teortaxesTex notes Xiaomi has unusually detailed public knowledge of its RL stack, from data construction to rollout characteristics to cost; what all this amounts to remains to be seen.
More from Models
- OpenAI's five biggest failures laid out: pretraining decline, alignment fears, enterprise lost to Claude — teortaxesTex · 2026-11-12
- Fable 5.2 Hits 41.3% on Remote Labor Index, Up From 2.5% a Year Ago — alliekmiller · 2026-11-12
- Beaker automates MIT's weight-space programming: one-neuron edit lifts MATH-500 from 7% to 37% for under $4 — VictorSanh · 2026-11-12
- Researcher says Anthropic dodged questions on Mythos false positive rate; 37,966 findings, 2 exploited — OwariDa · 2026-11-12
- Grok spotted building meeting "Listen to Conversation" feature with live transcript — nima_owji · 2026-11-12
- MoonMath ships uncensored 763B DeepSeek-V4.1 via closed-form weight editing, claims zero capability loss — udmrzn · 2026-11-12