Chat about technology with Xiaomi and change life together.
Xiaomi's MiMo-V2.6-Pro AI model assists researchers in designing novel MOF materials that adsorb PFAS pollutants millions of times better than benchmarks, cutting R&D cycle from one month to 2-3 days through automated literature review, hypothesis generation, and computational simulation.
A team built Sensaro OS in 48 hours at Xiaomi's hackathon, using Xiaomi MiMo multimodal models to transform smart homes from command responders into intent-understanding environments that maintain candidate intents, assess risk, and coordinate digital twins with real devices through a structured Sensaro State.
Xiaomi's AI Hardware Innovation Team open-sources four experimental projects exploring how AI can integrate into daily life through hardware: a wearable memory accessory with local privacy, a desktop companion visualizing agent status, a personality-driven desktop robot with smart home control, and an expressive robotic lamp — all emphasizing wearability, privacy, visible collaboration, and lifelike interaction.
Xiaomi details the dual-system architecture behind Super XiaoAI: a fast/slow routing layer that cuts model latency 35% via no-think mode and RAG-Planning, plus a full-duplex dialogue engine with three-level context compression (60–86% token reduction) and four-type interruption handling for natural in-car conversation.
Xiaomi releases Xiaomi-TabLDM, a foundation model for tabular data that uses large-scale synthetic pretraining, efficient model scaling with dual-stream feature groups and sparse MoE, and test-time scaling to achieve top-tier performance on four public benchmarks and real-world industrial tasks without per-dataset retraining.
Xiaomi's Darwin Agent team, collaborating with Nanjing University and Tsinghua University, won the IJCAI-ECAI 2026 CAR-bench Challenge Open Track with their TRACE method, achieving a 70% stable pass rate by decomposing experience into reusable actions, comparing execution traces to locate errors, and retrieving context-relevant experience for in-car voice assistant reliability.
Xiaomi had eight papers accepted at EMNLP 2026, covering mobile GUI agents with self-evolving rewards and process reward agents, diffusion LLM training with informed masking, speculative decoding via early-layer collaboration, token-budgeted reasoning, neural architecture search with code semantics, and proactive mobile agents with pre-reasoning perception.
Xiaomi‑Robotics‑1 demonstrates that pre‑training on 100,000 hours of real‑world manipulation data and subsequent cross‑embodiment fine‑tuning yields a scalable robot policy model that improves with larger data and model sizes, achieves state‑of‑the‑art performance on multiple simulation benchmarks, and adapts efficiently to new tasks with minimal downstream data.
Daniel Povey, Xiaomi’s chief scientist for speech, was elected a 2026 ISCA Fellow in recognition of his groundbreaking work on acoustic modeling, sequence‑training methods such as Minimum Phone Error, and the creation of the Kaldi toolkit and its next‑generation extensions that power modern multilingual ASR and TTS systems.
Xiaomi‑Robotics‑U0, a 38‑billion‑parameter multimodal autoregressive model, unifies four embodied generation tasks, offers geometry‑preserving data augmentation, achieves an 83× inference speedup with FlashAR+, tops the WorldArena benchmark, and improves real‑robot OOD task completion by over 26%.
