Yankai Jiang

SkeMex enables medical agents to evolve their reasoning capabilities by transforming raw experience into structured, reusable skills, outperforming traditional memory systems.

Medical MLLMs can achieve state-of-the-art reasoning performance, rivaling trillion-parameter models, by learning interpretable latent spaces with visual-semantic supervision, even without increasing inference costs.

LLMs can now dynamically create and refine their own scientific tools at test time, outperforming agents stuck with static toolsets.