BioAlchemy: Distilling Biological Literature into Reasoning-Ready Reinforcement Learning Training Data

Create an open-source dataset pipeline that extracts reasoning-heavy Q&A from biology literature. Use this to fine-tune a small model for biological reasoning benchmarks.

Suggested repo: bio-reason-gen

"Train your models on actual scientific reasoning, not just textbook biology."

Estimated effort: 60h