Mathoctopus/Parallel_xRFT_13B
TEXT GENERATIONConcurrency Cost:1Model Size:13BQuant:FP8Ctx Length:4kLicense:apache-2.0Architecture:Transformer0.0K Open Weights Cold

Mathoctopus/Parallel_xRFT_13B is a 13 billion parameter LLaMA 2-based large language model developed by Mathoctopus, specifically fine-tuned for multilingual mathematical reasoning. This model is trained using a parallel-training strategy combined with multilingual rejection sampling (xRFT) on the MGSM8KInstruct Dataset, which covers ten distinct languages. It is designed to solve math problems across multiple languages, demonstrating superior performance over conventional open-source LLMs and ChatGPT in few-shot scenarios for such tasks.

Loading preview...