Xiaomi's MiMo-V2.6 tech report: ~7k RL training data open-sourced, model shipped within a week of final RL run
rbhar90 · x · 2026-09-23
Xiaomi released the MiMo-V2.6 omni-modal tech report, centering on scaled RL toward self-improvement: asynchronous training consumes 1,568 samples and 2.73.7B tokens per step at up to 1M context; RL environments span code, general, visual, and cyber domains; groupwise agentic grading yields better reward signals, with a frozen MoE router and multi-layer defenses against reward hacking. The team plans to open-source 7k RL training samples plus the environments and framework, and shipped the model and report less than a week after starting the final RL run.
More from Models
- HackerGLM games the grader: guesses 11 hidden test IDs to pass a buggy RL environment — echen · 2026-11-12
- ADM 3 Cloud flops in image evals: ranks #86 of 107 in text-to-image — ArtificialAnlys · 2026-11-12
- YC-backed Monte launches post-training platform to build proprietary models from your agents' real work — ycombinator · 2026-11-12
- Leak Claims OpenAI Will Launch Entire GPT-6.2 Family Tomorrow to Counter Opus 5.5 and Fable 5.2 — heypearlai · 2026-11-12
- ARC-AGI-3 leaderboard: models score 90%+ but at $10K+ per task — haider1 · 2026-11-12
- LMArena Launches Agent Arena for Head-to-Head Agent Evaluation — arena · 2026-11-12