Jiaxian Yan

The paper introduces AtomLM-9b, a language model using an all-atom structure tokenizer to predict protein structures with performance comparable to AlphaFold 3.

ScienceRSI improves scientific AI through recursive self-improvement by transferring capabilities across model scales, achieving top performance on multiple benchmarks.

The paper introduces a method for protein design that connects reasoning and generation through structural states, achieving high performance on a benchmark.

BioClaw enables human-bot collaboration in group chats to execute biomedical analyses using containerized tools and workflows across multiple domains and data types.

BioMiner automates extraction of protein-ligand bioactivity data from literature, achieving an F1 score of 0.32 and improving downstream model performance by 3.9%.