daVinci-origin-7B

daVinci-origin-7B is a fully transparent, 7-billion parameter foundation model trained from scratch. It serves as a "clean-room" baseline for the research paper Data Darwinism -- Part1: Unlocking the Value of Scientific Data for Pre-training.

Unlike most open-source models, daVinci-origin-7B was explicitly trained on a dataset strictly excluding scientific content (books and research papers). This unique design allows researchers to unambiguously attribute performance gains to specific domain data injection strategies during continued pre-training.