# InternAgent

A unified agentic framework for long-horizon, autonomous scientific discovery, covering hypothesis generation through automated experimental execution across physical, biology, earth and life science domains.

InternAgent is a framework for end-to-end, autonomous scientific research, supporting both algorithm-discovery tasks and empirical, dry- or wet-lab-style discovery. It reports leading results on scientific reasoning benchmarks such as GAIA, HLE, GPQA and FrontierScience, and its scope spans twelve types of research tasks, from generating a hypothesis to reproducing published papers and running deep research on complex problems.

Its solution-optimization subsystem, MLEvolve, is released separately as the open-source method that reached first place on MLEBench for algorithm-design tasks. The project installs with conda and a `requirements.txt` file, needs API keys configured in a `.env` file, and was previously named NovelSeek before being renamed InternAgent.

Website: https://discovery.intern-ai.org.cn/home