Objective: Clinical entity linking is the task of associating spans in clinical text with unique concepts from standardized clinical ontologies. It is especially important in the clinical domain, given the complexity of terminology and the critical need for accurate interpretation by machine learning algorithms. Our goal is to survey methods for clinical entity linking across languages over the past three years, identify current trends in methodology, and analyze the impact of Large Language Models (LLMs) on solving this task. Methods: We queried multiple paper databases for papers on clinical entity linking and normalization published between 2023 and 2025. We extracted and categorized the system and model architectures used in the papers, as well as the datasets, languages, entity types, and ontologies used for method validation. Results: We analyzed 43 studies and found that the majority of methods use a retrieve-and-rerank architecture and employ transformer encoder models, with an increasing number of papers incorporating decoder models into the pipeline. We identified benchmark datasets for the task and the current state-of-the-art methods for each dataset. We discuss the multilingual strategies used by authors, including the use of multilingual models and neural machine translation from low-resource languages to high-resource languages such as English. Conclusion: LLMs have been successfully applied to a wide range of tasks; however, applying them directly to entity linking has encountered challenges due to the large search space of clinical concepts. Incorporating LLMs into the entity linking pipeline has proven beneficial for candidate reranking and clinical term standardization. The lack of labeled datasets in different languages, as well as the complexity of clinical terms, remains a challenge for clinical entity linking.

News articles often include repeated messages pushing for a particular viewpoint on a topic. These narratives can be explicit or hidden and appear in different articles, frequently leading to the spread of disinformation. Commonly, the approaches for narratives classification rely on custom, domain-specific taxonomies. In literary theory, though, there are well established frameworks for classification of narratives at generic level. We put forward the hypothesis that applying these existing frameworks in the area of narrative classification might provide a number of benefits, e.g. classification at a generic level might be useful by itself or as a feature for training on a specialized taxonomy. To this end, we propose a Generalized Annotation Scheme. The annotation scheme classifies narratives according to established literary frameworks, specifically Booker’s seven basic plot types, Reagan’s six emotional arcs, and the six conflict categories. We created a dataset of news article paragraphs, based on SemEval 2025 Task 10, Subtask 2. We annotated the paragraphs in the dataset along the three dimensions of our annotation scheme. Furthermore, we explore automatic annotation conducted using recent advanced language models, both proprietary and open-source. Our results indicate that employing state-of-the-art commercial LLMs for annotation tasks is feasible, paving the way for scalable, consistent, and cost-effective narrative analysis tasks.

SemEval-2026 Task 13 investigates machine-generated code detection across multiple programming languages and application scenarios, asking participating systems to generalize to unseen languages and domains. This paper describes our participation in Subtask A (binary classification) and explores both pretrained code encoders and lightweight feature-based methods. We design ratio-based features that are less sensitive to snippet length. To support the extraction of descriptiveness-related signals, we use parsing engines and a programming-language classifier. Additionally, we train a separate code-vs-text line classifier to identify raw natural language segments embedded within samples. We combine a shallow decision tree with heuristic rules derived from data analysis to produce the final predictions. Our approach is computationally efficient, requires only CPU resources for training, and achieves near-instant inference time, offering a lightweight alternative to large pretrained models.

This paper introduces a novel two-step multi-class classification system to identify varying degrees of machine involvement in Bulgarian text. As Large Language Models (LLMs) proliferate, distinguishing original human writing from machine-assisted or machine-generated content is crucial to prevent misinformation and preserve educational integrity. We developed a comprehensive dataset in Bulgarian encompassing purely human-written texts and four distinct mixed-content categories, such as machine-continued and machine-polished text. Our proposed pipeline utilises a binary ensemble classifier combining stylometric features, XLM-RoBERTa, and an adapted Binoculars model to first filter out human-written text. Subsequently, a specialised multi-class Support Vector Machine categorises the remaining machine-involved texts. Our findings indicate that this hierarchical approach improves the reliable identification of human authorship and enhances the overall accuracy of multi-class text detection compared to single-step methods.

The paper presents an approach for the recognition of toxic habits named entities in Spanish clinical texts. The approach was developed for the ToxHabits Shared Task. Our team participated in subtask 1, which aims to detect substance use and abuse mentions in clinical case reports and classify them in four categories (Tobacco, Alcohol, Cannabis, and Drug). We explored various methods of utilizing LLMs for the task, including zero-shot, few-shot, and prompt optimization, and found that GPT-4.1's few-shot prompting performed the best in our experiments. Our method achieved an F1 score of 0.65 on the test set, demonstrating a promising result for recognizing named entities in languages other than English.

Multi-step symbolic reasoning is essential for robust financial analysis; yet, current benchmarks largely overlook this capability. Existing datasets such as FinQA and ConvFinQA emphasize final numerical answers while neglecting the intermediate reasoning steps required for transparency and verification. To address this gap, we introduce FinChain, the first benchmark specifically designed for verifiable Chain-of-Thought evaluation in finance. FinChain spans 58 topics across 12 financial domains, each represented by parameterized symbolic templates with executable Python code that enable fully machine-verifiable reasoning and scalable, contamination-free data generation.To assess reasoning capacity, we propose ChainEval, a dynamic alignment measure that jointly evaluates both the final-answer correctness and the step-level reasoning consistency. Our evaluation of 26 leading LLMs reveals that even frontier LLMs exhibit clear limitations in symbolic financial reasoning, while domain-adapted and math-enhanced fine-tuned models can substantially narrow this gap.Overall, FinChain exposes persistent weaknesses in multi-step financial reasoning and provides a foundation for developing trustworthy, interpretable, and verifiable financial AI. This project is available at https://github.com/mbzuai-nlp/finchain.git.

Large language models are increasingly deployed across professional domains, bringing hard-to-predict risks, including the generation of harmful or disrespectful content. Although substantial progress has been made in developing safety evaluation datasets, existing resources remain overwhelmingly English- and Chinese-centric. This limitation is particularly pronounced when evaluating languages that operate within shared sociocultural, legal, and ethical contexts. To address this gap, we introduce Schützen: a German–Bulgarian safety dataset designed to assess model answerability under risk, covering both a low-resource language (Bulgarian) and a high-resource language (German). Experiments with multilingual and language-specific LLMs reveal pronounced cross-language differences in safety behavior, highlighting the necessity of tailored, region-specific evaluation resources to support the responsible deployment of LLMs in Germany and Bulgaria. Datasets and code are available at https://github.com/xnlp-lab/Schutzen. Warning: this paper contains examples that may be offensive, harmful, or biased.

Reliable evaluation of open-ended question answering remains a bottleneck for measuring answer correctness of modern LLMs. Unlike multiple-choice tasks, free-form answers may be correct in many surface forms and may fail in qualitatively different ways, including incompleteness, contradiction, overgeneration, and endorsement of false premises. Existing judgment-based and similarity-based metrics often collapse these distinctions. We address this gap with three reusable contributions. First, we introduce a semantic correctness taxonomy that assigns open-ended answers to eight ordered classes, separating verbose-but-correct answers from those contaminated by hallucinated content. Second, we release CAP-Correctness, an 8.8k-example benchmark spanning widely used QA datasets, and CAP-Statements, an 11k-example dataset for converting question-answer pairs into declarative statements for natural language inference (NLI) training and statement-based evaluation. Third, we introduce CAP (Context-Aware Precision), a reference-based metric that scores question-conditioned statements using bidirectional NLI. Under a monotonicity protocol testing whether metrics respect the taxonomy's intended ordering, CAP outperforms established baselines.

FinMMEval 2026 Task 2 evaluates short-answer financial question answering over multilingual evidence. Each final-test item pairs an English question with financial statements and news in English, Chinese, Japanese, Spanish, and Greek. Participating systems submit one concise answer per item in JSONL format. The final-test set contains 256 items, split evenly between easy and expert tiers; each tier contains four question templates instantiated over 32 company-report groups. Gold answers were withheld during submission, and systems were ranked by macro-averaged item-level ROUGE-1 F1 against organizer-held reference answers. The final leaderboard includes 12 ranked submissions. The strongest systems are closely clustered, with the top four separated by less than one percentage point in ROUGE-1 F1. The submitted system papers document retrieval-augmented generation, cross-lingual evidence handling, structured prompting, answer compression, and validation strategies.

As social media continues gaining a significant foothold in our everyday lives, it has become a potential channel for spreading propaganda. One way to address this problem is to identify the use of persuasion techniques in text. It is a challenging task because such techniques are complex making them difficult to recognize while reading. The paper presents an approach for detecting persuasion techniques used in text. We focus our work on a Bulgarian corpus, composed of news articles and parliament debates, both accurately annotated with 23 persuasion techniques. We provide insight into existing Bulgarian language models by evaluating them on this subjective task. For this purpose, we conduct experiments involving three paradigms for a multinomial classification task: (i) a traditional ML model, (ii) a fine-tuned transformer-based model, and (iii) a prompt-based LLM. We use different approaches, including two levels of distinction: the classification level (paragraph or sentence) and the type of classification (binary or multiple). Our analysis of the results showed that monolingual state-of-the-art models outperform their multilingual counterparts on low-resource languages such as Bulgarian.

We present the setup and the tasks of the FinMMEval Lab at CLEF 2026, which introduces the first multilingual and multimodal evaluation framework for financial Large Language Models (LLMs). While recent advances in financial natural language processing have enabled automated analysis of market reports, regulatory documents, and investor communications, existing benchmarks remain largely monolingual, text-only, and limited to narrow subtasks. FinMMEval 2026 addresses this gap by offering three interconnected tasks that span financial understanding, reasoning, and decision-making: Financial Exam Question Answering, Multilingual Financial Question Answering (PolyFiQA), and Financial Decision Making. Together, these tasks provide a comprehensive evaluation suite that measures models' ability to reason, generalize, and act across diverse languages and modalities. The lab aims to promote the development of robust, transparent, and globally inclusive financial AI systems, with datasets and evaluation resources publicly released to support reproducible research.

This paper presents a transformer-based approach to solving the SympTEMIST named entity recognition (NER) and entity linking (EL) tasks. For NER, we fine-tune a RoBERTa-based (1) token-level classifier with BiLSTM and CRF layers on an augmented train set. Entity linking is performed by generating candidates using the cross-lingual SapBERT XLMR-Large (2), and calculating cosine similarity against a knowledge base. The choice of knowledge base proves to have the highest impact on model accuracy.

FinMMEval 2026 Task 1 evaluates multilingual financial multiple-choice question answering in English, Chinese, Arabic, and Hindi. The task tests whether systems can select the correct answer to finance questions involving domain terminology, numerical interpretation, and conceptual financial reasoning across languages and scripts. The final-test set contains 800 questions, with 200 questions per language; gold answers were withheld during submission, and each language was ranked independently by accuracy. The final leaderboards contain 13 English, 11 Chinese, 11 Arabic, and 10 Hindi ranked submissions. Top accuracies range from 92.0

The usage of memes and other visual material coupled with text on social media has been on the rise recently. Recognizing that visual signals are consumed quickly and can trigger emotional responses. It has become essential to discern the sentiment of such content, as it could significantly influence social media users. The paper focuses on the sentiment of memes on popular social networking platforms such as Instagram, Reddit, Facebook, and Tumblr. Our goal is to understand how these memes affect people in a positive, negative, or neutral way. We create a balanced dataset of 5,592 memes using distant supervision, i.e., automatically assigning sentiment labels based on different social media attributes, e.g., hashtags. We verify the accuracy of these labels by manually checking a random subset of the data. We conduct unimodal and multimodal experiments to explore how different cues contribute to identifying sentiment. Our results show that multimodal approaches, combining images and text, effectively identify the emotions in memes. We further experiment with novel closed and open-source LLMs, and we show that they outperform traditional multimodal approaches. The dataset is released publicly.

We present SlavicNLP 2025 Shared Task on Detection and Classification of Persuasion Techniques in Parliamentary Debates and Social Media. The task is structured into two subtasks: (1) Detection, to determine whether a given text fragment contains persuasion techniques, and (2) Classification, to determine for a given text fragment which persuasion techniques are present therein using a taxonomy of 25 persuasion technique taxonomy. The task focuses on two text genres, namely, parliamentary debates revolving around widely discussed topics, and social media, in five languages: Bulgarian, Croatian, Polish, Russian and Slovene. This task contributes to the broader effort of detecting and understanding manipulative attempts in various contexts. There were 15 teams that registered to participate in the task, of which 9 teams submitted a total of circa 220 system responses and described their approaches in 9 system description papers.

The linking of clinical entities is a crucial part of extracting structured information from clinical texts. It is the process of assigning a code from a medical ontology or classification to a phrase in the text. The International Classification of Diseases - 10th revision (ICD-10) is an international standard for classifying diseases for statistical and insurance purposes. Automatically assigning the correct ICD-10 code to terms in discharge summaries will simplify the work of healthcare professionals and ensure consistent coding in hospitals. Our paper proposes an approach for linking clinical terms to ICD-10 codes in different languages using Large Language Models (LLMs). The approach consists of a multistage pipeline that uses clinical dictionaries to match unambiguous terms in the text and then applies in-context learning with GPT-4.1 to predict the ICD-10 code for the terms that do not match the dictionary. Our system shows promising results in predicting ICD-10 codes on different benchmark datasets in Spanish - 0.89 F1 for categories and 0.78 F1 on subcategories on CodiEsp, and Greek - 0.85 F1 on ElCardioCC.

ImageCLEF has been a part of CLEF (Conference and Labs of the Evaluation Forum) for more than 20 years. Started in 2003, ImageCLEF is an evaluation initiative which promotes the evaluation of technologies for annotation, indexing, retrieval, or generation of multimodal data. It provides access to large amounts of challenging data in very diverse use cases like medicine, argumentation, reasoning, generation, or content recommendation. In its 23rd edition, ImageCLEF will have four main tasks: (i) a Medical task involving concept detection and caption prediction in radiology images, synthetic medical images created with Generative Adversarial Networks (GANs), Visual Question Answering for improving the diagnosis and classification of real medical gastrointestinal images, and multimodal dermatology response generation, (ii) a joint ImageCLEF-Touché task Image Retrieval/Generation for Arguments to convey the premise of an argument, (iii) the ToPicto task which involves converting either text or speech into a meaningful sequence of pictograms and (iv) a new Multimodal Reasoning task addressing question answering and reasoning generation. In its last edition in 2024, 90 users and 31 unique teams submitted runs, totaling 257 runs, revealing a good impact in the community, similar to previous years.