Back

Research Papers

Our published and upcoming research papers on multilingual language models, reasoning alignment, and related topics.

Noema-T1: A Multilingual Model Family for Enhanced Reasoning

Aayan Mishra, Ameer Azam, Arush Sharma, Jordan Pfost and Hafedh Hichri

June 2025
Coming Soon

Abstract

We present Noema-T1, a family of multilingual language models trained on 100+ languages and finetuned for logic, truthfulness, and reflection. Our approach combines GPRO alignment techniques with efficient training methods to create models that can reason effectively across cultural and linguistic boundaries.

Keywords

Multilingual LLMs
Reasoning
GPRO
QLoRA

Citation Guidelines

If you use our models or research in your work, please cite the relevant paper using the following format:

@article{noema2025t1,
title={Noema-T1: A Multilingual Model Family for Enhanced Reasoning},
author={Aayan Mishra, Azam Ameer, Arush Sharma, Hafedh Hichri and Jordan Pfost},
journal={arXiv preprint},
year={2025}
}