While Large Language Models (LLMs) are known for their In-Context Learning (ICL) capabilities, there is no consensus on the underlying mechanisms. A key point of debate is whether ICL allows models to adapt to unseen tasks without parameter updates—that is, whether they can extrapolate. In this study, we address this question by constructing an arithmetic dataset based on the bivariate linear function z=ax+by to train a model and quantitatively evaluate its interpolation and extrapolation abilities through ICL. Our results show that while extrapolation was not achieved within our experimental design, tasks that were partially learned could be solved. We also found that the model acquires internal representations that can distinguish unseen tasks, and that greater task diversity in the training dataset improves ICL capabilities.

In this paper we describe a Polish news corpus as an attempt to create a filtered, organized and representative set of texts coming from contemporary online press articles from two major Polish TV news providers: commercial TVN24 and state-owned TVP Info. The process consists of web scraping, data cleaning and formatting. A random sample was selected from prepared data to perform a classification task. The random forest achieved the best prediction results out of all considered models. We believe that this dataset is a valuable contribution to existing Polish language corpora as online news are considered to be formal and relatively mistake-free, therefore, a reliable source of correct written language, unlike other online platforms such as blogs or social media. Furthermore, to our knowledge, such corpus from this period of time has not been created before. In the future we would like to expand this dataset with articles coming from other online news providers, repeat the classification task on a bigger scale, utilizing other algorithms. Our data analysis outcomes might be a relevant basis to improve research on a political polarization and propaganda techniques in media.

This paper investigates the effectiveness of automatic annotator assignment for text annotation in expert domains. In the task of creating high-quality annotated corpora, expert domains often cover multiple sub-domains (e.g. organic and inorganic chemistry in the chemistry domain) either explicitly or implicitly. Therefore, it is crucial to assign annotators to documents relevant with their fine-grained domain expertise. However, most of existing methods for crowdsoucing estimate reliability of each annotator or annotated instance only after the annotation process. To address the issue, we propose a method to estimate the domain expertise of each annotator before the annotation process using information easily available from the annotators beforehand. We propose two measures to estimate the annotator expertise: an explicit measure using the predefined categories of sub-domains, and an implicit measure using distributed representations of the documents. The experimental results on chemical name annotation tasks show that the annotation accuracy improves when both explicit and implicit measures for annotator assignment are combined.

It is known that word embeddings exhibit biases inherited from the corpus, and those biases reflect social stereotypes. Recently, many studies have been conducted to analyze and mitigate biases in word embeddings. Unsupervised Bias Enumeration (UBE) (Swinger et al., 2019) is one of approach to analyze biases for English, and Hard Debias (Bolukbasi et al., 2016) is the common technique to mitigate gender bias. These methods focused on English, or, in smaller extent, on Indo-European languages. However, it is not clear whether these methods can be generalized to other languages. In this paper, we apply these analyzing and mitigating methods, UBE and Hard Debias, to Japanese word embeddings. Additionally, we examine whether these methods can be used for Japanese. We experimentally show that UBE and Hard Debias cannot be sufficiently adapted to Japanese embeddings.

In this paper we present two methods for automatic common sense knowledge evaluation for Japanese entries in ConceptNet ontology. Our proposed methods utilize text-mining approach: one with relation clue words and WordNet synonyms, and one without. Both methods were tested with a blog corpus. The system based on our proposed methods reached relatively high precision score for three relations (MadeOf, UsedFor, AtLocation), which is comparable with previous research using commercial search engines and simpler input. We analyze errors and discuss problems of common sense evaluation, both manual and automatic and propose ideas for further improvements.