verify-ppt/marin_100b_7b

Synthetic Pre-pretraining Survives Scale, but Not as a Grammatical Prior

This model is part of the study in the paper Synthetic Pre-pretraining Survives Scale, but Not as a Grammatical Prior. It is a SmolLM3-based causal language model trained with a pre-pretraining and pre-training pipeline as described in the paper.

For code and usage instructions, please refer to the GitHub repository. Additional project resources are available on the Hugging Face organization page.