Bilingual Evaluation of Language Models on General Knowledge in University Entrance Exams with Minimal Contamination
This work addresses the need for bilingual evaluation benchmarks for language models, particularly in educational contexts, but it is incremental as it builds on existing datasets like MMLU.
The authors tackled the problem of evaluating language models on bilingual general knowledge from university entrance exams, creating the UNED-ACCESS 2024 dataset with 1003 questions in Spanish and English. Results showed that reasoning questions are challenging, smaller models degrade up to 37% more in Spanish than English, and model rankings correlate highly (0.98 Pearson) with MMLU, indicating the dataset's representativeness.
In this article we present UNED-ACCESS 2024, a bilingual dataset that consists of 1003 multiple-choice questions of university entrance level exams in Spanish and English. Questions are originally formulated in Spanish and translated manually into English, and have not ever been publicly released. A selection of current open-source and proprietary models are evaluated in a uniform zero-shot experimental setting both on the UNED-ACCESS 2024 dataset and on an equivalent subset of MMLU questions. Results show that (i) reasoning questions are challenging for models, (ii) smaller models perform worse than larger models and degrade faster in Spanish than in English and (iii) the performance gap between languages is negligible for the best models and grows up to 37% for smaller models. Model ranking on UNED-ACCESS 2024 is almost identical in English and Spanish, and has also a high correlation (0.98 Pearson) with ranking on MMLU, suggesting that a small dataset is sufficiently diverse and representative to measure performance by discipline.