RePo-OLMo2-7B-stage2-L10
This repository contains a 1B parameter base model continually-pretrained from allenai/OLMo-2-1124-7B on 50B tokens using the context Re-Positioning (RePo) technique introduced in our paper RePo: Language Models with Context Re-Positioning.
Model Details
- Developed by: Sakana AI
- Model type: Autoregressive Language Model
- License: Apache License, Version 2.0
- Paper: https://arxiv.org/abs/2512.14391
- Code: https://github.com/SakanaAI/repo
Model Description
This model is based on allenai/OLMo-2-1124-7B model, which has completed its Stage-1 pre-training. We continually pre-trained it on 50B tokens from the Stage-2 allenai/dolmino-mix-1124 dataset. Evaluation was conducted using an adapted version of olmes. Please refer to our repository and paper for details and results.
Usage
This model is provided for research and development purposes only and should be considered as an experimental prototype. It is not intended for commercial use or deployment in mission-critical environments. Use of this model is at the user's own risk, and its performance and outcomes are not guaranteed. Sakana AI shall not be liable for any direct, indirect, special, incidental, or consequential damages, or any loss arising from the use of this model, regardless of the results obtained. Users must fully understand the risks associated with the use of this model and use it at their own discretion.
Citation
@article{sakana2025repo,
title={RePo: Language Models with Context Re-Positioning},
author={Huayang Li, Tianyu Zhao, and Richard Sproat},
year={2025},
eprint={2512.14391},
archivePrefix={arXiv},
primaryClass={cs.LG},
url={https://arxiv.org/abs/2512.14391},
}