Senior Researcher, Microsoft Research
1 paper at NeurIPS 2025
We introduce a new comprehensive benchmark, MMTU, designed to evaluate models ability to understand, reason, and manipulate diverse tables.