GSM8k a dataset of 8.5K high quality linguistically diverse grade school math word problems.
Here we provide the Romanian translation of the GSM8k benchmark, translated with Systran.
This dataset is used as a benchmark and is part of the evaluation protocol for Romanian LLMs proposed in "Vorbeşti Româneşte?" A Recipe to Train Powerful Romanian LLMs with English Instructions (Masala et al., 2024)
Citation… See the full description on the dataset page: https://huggingface.co/datasets/OpenLLM-Ro/ro_gsm8k.
