Xinming Tu
University of Washington. Seattle.
Weiming Lake, Bo Ya Ta, Peking University
I am a Founding Member of Technical Staff at Phylo and a final-year PhD student at the Paul G. Allen School of Computer Science & Engineering at the University of Washington, advised by Sara Mostafavi.
My research centers on agentic science. I build rigorous benchmarks to understand how AI contributes to the research process and develop agentic architectures that make AI a better scientific collaborator—in a sense, “research about research.” Recent work includes BenchGuard (COLM 2026), BiomniBench, and DrugDiscoveryBench.
I also work on AI for regulatory genomics, developing deep learning methods for sequence-to-function prediction and multi-omics integration. Representative work includes SAGE-net (Nature Methods), CLUE (NeurIPS Oral), and DeepAllele (RECOMB 2026).
Before UW, I studied Biology and Computer Science at Peking University. Please see my CV for more information.
Selected Posts
Selected Publications
2026
- Benchmark of 82 expert-curated, verifiable early-stage drug-discovery tasks for evaluating coding agents across target identification, patent mining, structure-activity analysis, and lead optimization.