Being able to thoroughly assess massive multi-task language understanding (MMLU) capabilities is essential for advancing the applicability of multilingual language models. However, preparing such benchmarks in high quality native language is often costly and therefore limits the representativeness of evaluation datasets. While recent efforts focused on building more inclusive MMLU benchmarks, these are conventionally built using machine translation from high-resource languages, which may introduce errors and fail to account for the linguistic and cultural intricacies of the target languages. In this paper, we address the lack of native language MMLU benchmark especially in the under-represented Turkic language family with distinct morphosyntactic and cultural characteristics. We propose two benchmarks for Turkic language MMLU: TUMLU is a comprehensive, multilingual, and natively developed language understanding benchmark specifically designed for Turkic languages. It consists of middle- and high-school level questions spanning 11 academic subjects in Azerbaijani, Crimean Tatar, Karakalpak, Kazakh, Kyrgyz, Tatar, Turkish, Uyghur, and Uzbek. We also present TUMLU-mini, a more concise, balanced, and manually verified subset of the dataset. Using this dataset, we systematically evaluate a diverse range of open and proprietary multilingual large language models (LLMs), including Claude, Gemini, GPT, and LLaMA, offering an in-depth analysis of their performance across different languages, subjects, and alphabets. To promote further research and development in multilingual language understanding, we release TUMLU-mini and all corresponding evaluation scripts.

As the number of treebanks within the same language family continues to grow, the importance of establishing consistent annotation practices has become increasingly evident. In this paper, we evaluate various approaches to annotating Turkic postverbial constructions across UD treebanks. Our comparative analysis reveals that none of the existing methods fully capture the unique semantic and syntactic characteristics of these complex constructions. This underscores the need to adopt a balanced approach that can achieve broad consensus and be implemented consistently across Turkic treebanks. By examining the phenomenon and the available annotation strategies, our study aims to improve the consistency of Turkic UD treebanks and enhance their utility for cross-linguistic research.

This paper advances NLP research for the low-resource Uzbek language by evaluating two previously untested monolingual Uzbek BERT models on the part-of-speech (POS) tagging task and introducing the first publicly available UPOS-tagged benchmark dataset for Uzbek. Our fine-tuned models achieve 91% average accuracy, outperforming the baseline multi-lingual BERT as well as the rule-based tagger. Notably, these models capture intermediate POS changes through affixes and demonstrate context sensitivity, unlike existing rule-based taggers.

We present the first Universal Dependencies treebank for Uzbek, a low-resource language from the Turkic family. The treebank contains 500 sentences (5850 tokens) sourced from the news and fiction genres and it is annotated for lemmas, part-of-speech (POS) tags, morphological features, and dependency relations. We describe our methodology for building the treebank, which consists of a mix of manual and automatic annotation and discuss some constructions of the Uzbek language that pose challenges to the UD framework.