Scientific code repositories hold decades of human knowledge encoded in executable models, methods, and tools. According to a new paper posted on Hugging Face, converting that knowledge into reliable learning experience is difficult because of fragmented toolchains, implicit domain conventions, and specialized correctness criteria — a problem the authors call the scientific experience bottleneck.
To address it, the researchers introduce ScienceIDE, infrastructure designed to turn the world's scientific code into programmable environments for scientific agents. Guided by expert-defined scientific cases and acceptance criteria, agents transform repositories into executable environments that support task generation, execution, and scientific verification.
These environments are intended to serve as a shared foundation for supervised fine-tuning, reinforcement learning, and evaluation. Using verified interaction trajectories, the team trained three models: PhAI-IDE-72B, PhAI-IDE-9B, and PhAI-IDE-4B.
The model family reportedly shows gains in held-out scientific-code repair and across selected general-purpose benchmarks in code, reasoning, and knowledge, which the authors present as evidence of positive transfer from scientific experience to broader capabilities.
The paper positions ScienceIDE as a foundation for an integrated workspace for agent learning and scientific practice, making humanity's scientific software a shared substrate for developing scientific intelligence. Code and models are linked in the paper's repository and Hugging Face collection.