Tagged articles

large‑scale pretraining

4 articles · Page 1 of 1

Jul 16, 2026 · Artificial Intelligence

100k‑Hour “Plug‑and‑Play” Robot Base Model: Xiaomi‑Robotics‑1 Tests Scaling Laws

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.

Embodied AIScaling Lawslarge‑scale pretraining

0 likes · 11 min read

100k‑Hour “Plug‑and‑Play” Robot Base Model: Xiaomi‑Robotics‑1 Tests Scaling Laws

Machine Heart

Jul 15, 2026 · Artificial Intelligence

10,000‑Hour Human Data Powers the First World‑Action Model for Humanoids

Leveraging over 10,000 hours of human‑centric video and motion data, the Being‑M0.7 model introduces the first implicit world‑action system capable of full‑body mobile manipulation on humanoid robots, outperforming prior baselines across challenging tasks such as fish netting, mirror retrieval, and object transport.

Latent Action ModelVision-Motion MoTfull-body manipulation

0 likes · 17 min read

10,000‑Hour Human Data Powers the First World‑Action Model for Humanoids

Xiaomi Tech

Jun 12, 2026 · Artificial Intelligence

Building a Universal Sound Model: Xiaomi Dasheng’s 8‑GPU AI Engineering Journey

This article details Xiaomi Dasheng’s end‑to‑end approach to creating a universal sound representation model, from choosing a Masked Autoencoder pre‑training framework and scaling up to 1.2 B parameters on 300 TB of diverse audio, to six‑dimensional annotation, unified understanding‑generation architecture, and future audio‑scene generation.

Xiaomi Dashengaudio generationaudio representation

0 likes · 13 min read

Building a Universal Sound Model: Xiaomi Dasheng’s 8‑GPU AI Engineering Journey

Baidu Tech Salon

May 24, 2024 · Artificial Intelligence

HelixDock: A Large-Scale Pretrained Full-Atom Diffusion Model for Protein–Small Molecule Docking

HelixDock, a full‑atom diffusion model pretrained on a billion‑scale simulated docking dataset covering ~200,000 protein targets, delivers state‑of‑the‑art docking accuracy—85.6% success on PoseBusters and strong generalization on cross‑docking benchmarks—showing that massive data and model scaling dramatically improve AI‑driven drug discovery, and its code and data are fully open‑source.

AI for drug discoveryDeep LearningHelixDock

0 likes · 6 min read

HelixDock: A Large-Scale Pretrained Full-Atom Diffusion Model for Protein–Small Molecule Docking