Estonian Native Large Language Model Benchmark

TL;DR

This paper introduces a comprehensive Estonian LLM benchmark with diverse datasets, evaluating various models through human and LLM-based assessments, to advance Estonian NLP capabilities.

Contribution

It presents the first extensive Estonian LLM benchmark using native data, comparing multiple models and evaluation methods, including human and LLM judges.

Findings

Human scores correlate moderately to highly with benchmark results.

Claude 3.7 Sonnet as an LLM judge aligns well with human evaluations.

Open-source instruction-tuned models show competitive performance.

Abstract

The availability of LLM benchmarks for the Estonian language is limited, and a comprehensive evaluation comparing the performance of different LLMs on Estonian tasks has yet to be conducted. We introduce a new benchmark for evaluating LLMs in Estonian, based on seven diverse datasets. These datasets assess general and domain-specific knowledge, understanding of Estonian grammar and vocabulary, summarization abilities, contextual comprehension, and more. The datasets are all generated from native Estonian sources without using machine translation. We compare the performance of base models, instruction-tuned open-source models, and commercial models. Our evaluation includes 6 base models and 26 instruction-tuned models. To assess the results, we employ both human evaluation and LLM-as-a-judge methods. Human evaluation scores showed moderate to high correlation with benchmark evaluations, depending on the dataset. Claude 3.7 Sonnet, used as an LLM judge, demonstrated strong alignment with human ratings, indicating that top-performing LLMs can effectively support the evaluation of Estonian-language models.