Xiaomi LLM-Core Team
AI overview
Sign in and the AI will write an overview from our coverage.
Headlines · 1
- Xiaomi's MiMo-V2.6 report scales RL toward self-improvement
Xiaomi's LLM-Core Team released a technical report on the MiMo-V2.6 series, an omni-modal family that scales reinforcement-learning compute to push model intelligence. The report says it mid-trains on a broad multimodal corpus, then scales RL on a pretrained hybrid-SWA architecture with batches consuming 1,568 samples and 2.7–3.7B tokens per step at context lengths up to 1M, across code, general, visual and cyber environments, using groupwise agentic grading for better long-horizon reward signals. The team says it open-sourced training dynamics, RL environments and the RL framework, and froze the MoE router with multi-layer defenses against reward hacking.
Hugging Face · Papers · 🔥 0
Experience and discussion from the community
Share my Xiaomi LLM-Core Team experienceAsk about Xiaomi LLM-Core Team
Nobody has shared their experience with Xiaomi LLM-Core Team yet.