Alibaba Reportedly Leads $300 Million Funding for AI Benchmark Startup UniPat
Alibaba is reportedly leading a $300 million funding round at a $2.5 billion valuation for UniPat AI, an AI evaluation and benchmark training startup founded by former intern Li Kuan. Existing investors including Tencent Holdings and Sequoia are also participating in the deal. As human-generated internet data becomes scarce due to privacy and copyright restrictions, major tech companies are investing heavily in synthetic data infrastructure and evaluation tools. UniPat addresses both data scarcity and the issue of inflated benchmark scores that fail to reflect real-world model capabilities. UniPat provides realistic evaluation scenarios covering software engineering for coding agents, browser interaction tasks, and visual reasoning for multimodal models. Founder Li Kuan previously worked on post-training analysis, synthetic data, and reinforcement learning at Alibaba's Tongyi AI Lab.
## BACKGROUND
AI training traditionally relies on massive web-scraped datasets, but data contamination and benchmark gaming often lead to models scoring high on standardized tests while underperforming in production. Synthetic data involves artificially generating realistic training samples using AI algorithms, allowing developers to safely scale datasets without privacy or copyright conflicts.