Skip to content
Rath Lab

Make Everything Agentic

Rath Lab is an AI research lab at Shanghai Lynwe Technology. We study the core capabilities of agents and harnesses.

Agent = Model + Harness

Progress in the harness translates directly into more capable agents. Our research complements the direction foundation-model developers explore: foundation-model companies build models, adapt them to a domain by fine-tuning, and explore the ceiling of model capability; we build agents and harnesses, adapt them to a domain by redesigning the harness, and explore how to organize today’s models into agents that stay reliable over long horizons.

  • arXiv

    OpenRath: Session-Centered Runtime State for Agent Systems

    Fukang Wen, Zhijie Wang, Ruilin Xu

Our work

View all

Research directions

All directions

Make everything agentic, with us