news
BenchGuard: Who Guards the Benchmarks? is accepted to COLM 2026!
New preprint from Scale Labs x Phylo! DrugDiscoveryBench: Can Coding Agents Assist Early-Stage Drug Discovery? — a benchmark of 82 expert-curated drug-discovery tasks evaluating how reliably coding agents handle multi-step computational and information-retrieval workflows.
New preprint on bioRxiv! BiomniBench: Process-level Evaluation of LLM Agents for Real-world Biomedical Research — a process-level benchmark that scores full agent trajectories against expert-designed rubrics on 100 data-analysis tasks from top-tier biomedical papers.
New preprint on bioRxiv! A modality gap in personal-genome prediction by sequence-to-function models is now available.
Two papers recently accepted! SAGE-net at Nature Methods and DeepAllele at RECOMB 2026.
I posted my third blog - AI for General Science - Large language models for scientific hypothesis/research ideas generation
This summer, I will join in the Aviv Regev’s Group at Genentech as a Research Intern, work with Romain Lopez.
CLUE is accepted to NeurIPS 2022 as Oral presentation! See you in New Orleans!
I posted my second blog - Collection of some resources related to the PhD journey
I gave a talk about our CLUE method on the 9th Winter q-bio conference!
I posted my first blog - Seurat CCA? It’s just a simple extension of PCA!
Our method CLUE is the winner in all categories of Match Modality task in Multimodal Single-Cell Data Integration Competition(NeurIPS Competition 2021) 
