A Chinese handbook for the messy reality of VLA robots

What it does

VLA Handbook is a Chinese-language, engineering-first knowledge base for algorithm engineers entering the Vision-Language-Action robotics field. It houses roughly 380 theoretical documents across ten topics, plus curated “field notes” distilled from 300+ Chinese social media posts, 165 English community sources, and 47 high-traffic GitHub Issues. The goal is to close the gap between understanding a paper and getting its code to survive contact with real hardware.

The interesting bit

Rather than pretending to be the only source you need, the README openly concedes what WeChat articles, Awesome lists, and Twitter threads do well—then stakes its claim on the dirty work: multimodal sync, action-space alignment, and hardware-specific training failures. An auto-updating pipeline scores new papers daily, distills community chatter, and even tracks robotics-industry funding and product launches, turning a static repo into a living intelligence feed.

Key highlights

• 380+ theory docs with entry scripts, key hyperparameters, and tensor-shape sanity checks
• Curated field notes from Xiaohongshu, Hugging Face blogs, LeRobot Discord, and GitHub Issues covering real deployment failures
• Daily auto-pipeline that scores papers, generates deep parses, and tracks 15 method families with trend charts
• Industry radar monitoring robotics companies for funding, products, IPOs, and partnerships
• Permanently versioned in Git with RSS feeds and OPML subscriptions for tracking the firehose

Verdict

Worth bookmarking if you read Chinese and are trying to move VLA research from Jupyter notebook to physical robot. Skip it if you need a gentle English-language primer or a small, static reference.