Ro GSM8K is a Romanian translation of the GSM8K benchmark, containing 8.5 thousand grade school math word problems. The dataset was translated with Systran and is authored by OpenLLM-Ro. It was last updated on September 25, 2024, and serves as part of the evaluation protocol for Romanian large language models proposed in the research paper "Vorbeşti Româneşte?" A Recipe to Train Powerful Romanian LLMs with English Instructions.
Use Cases
- Benchmarking the mathematical reasoning capabilities of Romanian language models based on the translated word problems.
- Training models for multilingual math problem-solving tasks using the Romanian-language content.
- Evaluating the quality of machine translation for educational and technical text via the Systran-translated problems.
Strengths
- Contains 8.5 thousand high-quality, linguistically diverse math problems.
- Serves as a dedicated benchmark for evaluating Romanian language models, as cited in a 2024 research paper.
Limitations
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count is unknown, which may limit suitability assessment.
- Description metadata is limited; actual data quality requires manual inspection after download.
Provenance
- Source
- OpenLLM-Ro, based on the original GSM8K dataset.
- Collection Method
- Translated from the original English GSM8K benchmark using Systran.
- Freshness
- Last updated 2024-09-25 10:28:44; freshness should be verified.