LeiPhone WhaleX has learned that Jianxiong Li, the first doctoral graduate of Tsinghua’s Institute for AI Industry (AIR), has founded embodied intelligence company BASAL Intelligence, with a round led by FiveYuan Capital and participation from IVC, Linear Capital and WestSummit Capital. AIR associate professor Xianyuan Zhan serves as chief scientist, with a core team of seven long time collaborators.
Before forming the team, core members held top offers from ByteDance, Nvidia and Xiaomi and overseas PhD admissions, with combined abandoned pay packages near RMB 40 million; Li’s own highest package approached RMB 10 million. BASAL says it will not follow the mainstream pretrain base model plus post train deployment pattern, betting instead on an action native in-context embodied foundation model that adapts on site without retraining.
Today’s mainstream path pretrains a base model then post trains per task. It works in standard factories but breaks when environment, object or workflow changes, forcing fresh data and retraining. BASAL argues this is still industrial automation, not true embodied intelligence, and that language centred VLA routes have innate limits in fine manipulation and cross scene generalisation. Its goal is on site learning: faced with an unseen scene or task, the robot adapts in minutes from a few human demos and self trial.
BASAL’s method puts action, not language, at the centre, letting the model learn the relation between action, task feedback, failure correction and human demo, using long context to organise continuous physical interaction. Two core technologies back this: the cross embodiment model X-VLA, which with under 1,000 hours of pretraining topped third party benchmarks and won the IROS 2025 AgiBot challenge out of 400-plus teams; and ODEWorld, an efficient latent space world model for long horizon physical modelling.
Editor’s note: This is an adapted translation of the original LeiPhone report. The full source is at https://www.leiphone.com/category/ai/WbD0Tf6P1oA2WDdq.html (LeiPhone). It has been trimmed and restructured for readability for an international business audience.