StableAI (Tsinghua/Wenzhun)

foundation-model
research
automl
Published

October 1, 2024

Overview

StableAI is a research collaboration between Tsinghua University and Wenzhun Technology focused on building LimiX, a tabular foundation model that emphasizes training stability and robust generalization. The project bridges academic research and industry application in the Chinese AI ecosystem.

Key Product

LimiX is a tabular foundation model designed for stable and reliable predictions across diverse datasets. LimiX focuses on addressing common failure modes in tabular pre-training, such as negative transfer and distribution shift, through novel architectural and training innovations.

Team

The team brings together researchers from Tsinghua University, one of China’s top AI research institutions, and Wenzhun Technology, providing both academic rigor and industry perspective.

Outlook

As a research collaboration rather than a venture-backed startup, StableAI represents the academic frontier of tabular FMs in China. Their focus on training stability addresses a real technical challenge, and proximity to Tsinghua’s talent pipeline is a long-term asset.