SphereLab.

We believe the next leap in AI won't come from blindly stacking more parameters — it will come from finding the right principles. The simple ones. The ones that truly scale. So we hunt for them.

We work on the principles that shape how machines learn — the structures, symmetries, and quiet assumptions that decide what a model can generalize to. We believe the right idea is often worth more than another order of magnitude of compute, and we work to turn the ones we find into functional algorithms and systems.

How we work.  Curiosity over trends. Depth over speed. Insight before application.

Director.  Weiyang Liu
Members.  He Guo, Yangyi Huang, Zeju Qiu, Kexuan Shi, Siyuan Wang, Zhouliang Yu, Haoquan Zhang
Affiliated members.  Lijun Cai, Kailebo Jiefulisi, Jiale Kang, Hanxuan Li, Luocheng Liang, Siyuan Ma, Zihao Wang, Tim Z. Xiao, Sihan Yang