HuGME is an easy-to-use LLM assessment framework, which explicitly rates the Hungarian language skills and cultural knowledge of the models. It can be used to evaluate LLM outputs based on metrics such as hallucination, relevance of response, bias and more.

Educational content to help other researchers and students understand language models.

Hungarian Language Understanding Benchmark Kit. Includes evaluation scripts and a Python package for benchmarking models on HuLU.

e-magyar text processing system -- inter-module communication via tsv + REST API

A generic TSV-style format based intermodular communication framework and REST API implemented in Python

Conference and thesis projects of students

TSV files of the Parallel Bible Reader

Hungarian Choice of Plausible Alternatives Corpus

Hungarian version of the Stanford Sentiment Treebank

Hungarian version of the Recognising Textual Entailment datasets