Yevhen Kostiuk

Oxana Vitman, Yevhen Kostiuk, Grigori Sidorov et al.

Sarcasm detection is an essential task that can help identify the actual sentiment in user-generated data, such as discussion forums or tweets. Sarcasm is a sophisticated form of linguistic expression because its surface meaning usually contradicts its inner, deeper meaning. Such incongruity is the essential component of sarcasm, however, it makes sarcasm detection quite a challenging task. In this paper, we propose a model, that incorporates different features to capture the incongruity intrinsic to sarcasm. We use a pre-trained transformer and CNN to capture context features, and we use transformers pre-trained on emotions detection and sentiment analysis tasks. Our approach outperformed previous state-of-the-art results on four datasets from social networking platforms and online media.

Yevhen Kostiuk, Grigori Sidorov, Olga Kolesnikova

In natural language processing (NLP), lexical function is a concept to unambiguously represent semantic and syntactic features of words and phrases in text first crafted in the Meaning-Text Theory. Hierarchical classification of lexical functions involves organizing these features into a tree-like hierarchy of categories or labels. This is a challenging task as it requires a good understanding of the context and the relationships among words and phrases in text. It also needs large amounts of labeled data to train language models effectively. In this paper, we present a dataset of most frequent Spanish verb-noun collocations and sentences where they occur, each collocation is assigned to one of 37 lexical functions defined as classes for a hierarchical classification task. Each class represents a relation between the noun and the verb in a collocation involving their semantic and syntactic features. We combine the classes in a tree-based structure, and introduce classification objectives for each level of the structure. The dataset was created by dependency tree parsing and matching of the phrases in Spanish news. We provide baselines and data splits for each objective.

Yevhen Kostiuk, Atnafu Lambebo Tonja, Grigori Sidorov et al.

In this paper, we investigate the issue of hate speech by presenting a novel task of translating hate speech into non-hate speech text while preserving its meaning. As a case study, we use Spanish texts. We provide a dataset and several baselines as a starting point for further research in the task. We evaluated our baseline results using multiple metrics, including BLEU scores. The aim of this study is to contribute to the development of more effective methods for reducing the spread of hate speech in online communities.

Yevhen Kostiuk, Kenneth Enevoldsen, Peter Bjerregaard Vahlstrup et al.

Social norms reflect shared expectations on acceptable behavior. Measuring social norms alignment remains challenging, with existing approaches typically relying on artificial closed-form evaluations such as multiple-choice questionnaires or measuring agreement with predefined statements. In the context of this work, social norms alignment refers to measuring an agreement between solutions with respect to the social problem or dilemma. We propose a framework for measuring social norm alignment in naturalistic, free-form settings through solution matching. The framework enables us to measure alignment between any two dilemma responses e.g., LLMs to a human, LLMs to LLMs, or human to human. We introduce two metrics: stated and explicit agreement accuracy, and construct a dataset of 3k non-trivial social dilemmas in Danish. All dilemmas are assigned reference solutions derived from three panelists, who serve as culturally grounded judges. We evaluate the agreement of several LLMs and human responses in an interaction setup that resembles natural user-model conversations. Our results show that the proposed metrics produce consistent model rankings and reveal variation in agreement across different types of dilemmas, with higher agreement observed for topics such as neighbor conflicts and shared living situations. Overall, our work introduces a dataset and evaluation framework for studying culturally grounded social reasoning in naturalistic open-ended conversations.

Yevhen Kostiuk, Kenneth Enevoldsen

Instruction embedding models have become common among state-of-the-art models, however are evaluated using a single prompt per task. The single-point evaluation ignores a main problem of the instruction-based approach namely: sensitivity to the phrasing of the instruction. We present an empirical study of prompt sensitivity across 6 embedding models, 11 datasets, and 15 task-specific prompts per dataset, a total of 990. We show that reported scores misrepresent the distribution of scores over plausible prompts. The default prompt can both systematically understate or overstate performance. Furthermore, we show that the leaderboard ranking is not robust to prompt selection: by choosing prompts favorably, any model in our study can be promoted to first place. Our findings suggest that single-prompt evaluation is insufficient for instruction-tuned embedding models and that benchmarks should incorporate prompt robustness, either by evaluating over multiple prompts or by reporting sensitivity alongside point estimates.

Yevhen Kostiuk, Oxana Vitman, Łukasz Gagała et al.

In this work, we evaluated Lithuanian and general history knowledge of multilingual Large Language Models (LLMs) on a multiple-choice question-answering task. The models were tested on a dataset of Lithuanian national and general history questions translated into Baltic, Nordic, and other languages (English, Ukrainian, Arabic) to assess the knowledge sharing from culturally and historically connected groups. We evaluated GPT-4o, LLaMa3.1 8b and 70b, QWEN2.5 7b and 72b, Mistral Nemo 12b, LLaMa3 8b, Mistral 7b, LLaMa3.2 3b, and Nordic fine-tuned models (GPT-SW3 and LLaMa3 8b). Our results show that GPT-4o consistently outperformed all other models across language groups, with slightly better results for Baltic and Nordic languages. Larger open-source models like QWEN2.5 72b and LLaMa3.1 70b performed well but showed weaker alignment with Baltic languages. Smaller models (Mistral Nemo 12b, LLaMa3.2 3b, QWEN 7B, LLaMa3.1 8B, and LLaMa3 8b) demonstrated gaps with LT-related alignment with Baltic languages while performing better on Nordic and other languages. The Nordic fine-tuned models did not surpass multilingual models, indicating that shared cultural or historical context alone does not guarantee better performance.

Yevhen Kostiuk, Oxana Vitman, Łukasz Gagała et al.

In this work, we address the challenge of evaluating large language models (LLMs) on the short answer matching task for Latvian and Lithuanian languages. We introduce novel datasets consisting of 502 Latvian and 690 Lithuanian question-answer pairs. For each question-answer pair, we generated matched and non-matched answers using a set of alteration rules specifically designed to introduce small but meaningful changes in the text. These generated answers serve as test cases to assess the ability of LLMs to detect subtle differences in matching of the original answers. A subset of the datasets was manually verified for quality and accuracy. Our results show that while larger LLMs, such as QWEN2.5 72b and LLaMa3.1 70b, demonstrate near-perfect performance in distinguishing matched and non-matched answers, smaller models show more variance. For instance, LLaMa3.1 8b and EuroLLM 9b benefited from few-shot examples, while Mistral Nemo 12b underperformed on detection of subtle text alteration, particularly in Lithuanian, even with additional examples. QWEN2.5 7b and Mistral 7b were able to obtain a strong and comparable performance to the larger 70b models in zero and few shot experiments. Moreover, the performance of Mistral 7b was weaker in few shot experiments.

Artur Kiulian, Anton Polishko, Mykola Khandoga et al.

In this paper, we propose a model-agnostic cost-effective approach to developing bilingual base large language models (LLMs) to support English and any target language. The method includes vocabulary expansion, initialization of new embeddings, model training and evaluation. We performed our experiments with three languages, each using a non-Latin script - Ukrainian, Arabic, and Georgian. Our approach demonstrates improved language performance while reducing computational costs. It mitigates the disproportionate penalization of underrepresented languages, promoting fairness and minimizing adverse phenomena such as code-switching and broken grammar. Additionally, we introduce new metrics to evaluate language quality, revealing that vocabulary size significantly impacts the quality of generated text.