The deconstruction of gender stereotypes is essential to prevent discrimination, marginalization and gender-based violence. Despite the increasing attention to this issue, research in this field often focuses on explicitly sexist or hateful communication, leaving out all the cases where stereotypes are produced unconsciously or even with apparently positive intentions. Moreover, the identification and analysis of gender stereotypes is often a very subjective task, heavily influenced by the researcher’s background, beliefs and personal sensitivity. In this context GSI:detect, a dataset for gender stereotypes identification in Italian, has been annotated following a perspectivist approach that gives value to the different points of view of four annotators. It has been designed to address (i) the lack of resources focusing on naturally occurring and non-hateful language conveying implicit or ambiguous forms of gender stereotypes, and (ii) the scarcity of datasets that can capture multiple interpretations as well as the inherent variation and disagreement in human perception. Baseline experiments with several LLMs confirm the challenging nature and value of such a linguistic resource, revealing both apparent differences and limitations in performance among the evaluated models, and raising questions about the extent to which current LLMs are suitable for detection and classification tasks in this field. Content warning: Examples taken from the GSI:detect dataset may contain sensitive or potentially distressing content.

While proactivity, i.e., the ability to take the initiative and anticipate requests in order to improve the effectiveness of a conversation, has been traditionally investigated in task-oriented dialogues (e.g., booking a restaurant), less work addresses proactive behaviours in task-guidance dialogues (e.g., guide to execute recipes), where the expert instructor is supposed to interact and supervise a user in a real-world setting. We analyse a corpus of video-recorded task-guided dialogues and explore two key features of proactivity in this context: (i) the impact of multimodal features, with respect to chat-based dialogues; (ii) the impact of instructions and actions grounded in a real situation. Through a comparison between task-oriented and task-guidance annotated dialogues, we find that task-guided dialogues are highly collaborative interactions, where preventing mistakes and maintaining the correct process order is essential for achieving the dialogue goal. In addition, the video information available in the task-guidance setting can be corrective for false positive proactive behaviours, although without introducing substantial differences. To support our analysis and to foster further research we provide a corpus of multimodal task-guidance dialogues annotated according to proactivity.

Large Language Models (LLMs) consistently excel in diverse medical Natural Language Processing (NLP) tasks, yet their substantial computational requirements often limit deployment in real-world healthcare settings. In this work, we investigate whether “small” LLMs (around one billion parameters) can effectively perform medical tasks while maintaining competitive accuracy. We evaluate models from three major families—Llama-3, Gemma-3, and Qwen3—across 20 clinical NLP tasks among Named Entity Recognition, Relation Extraction, Case Report Form Filling, Question Answering, and Argument Mining. We systematically compare a range of adaptation strategies, both at inference time (few-shot prompting, constraint decoding) and at training time (supervised fine-tuning, continual pretraining). Fine-tuning emerges as the most effective approach, while the combination of few-shot prompting and constraint decoding offers strong lower-resource alternatives. Our results show that small LLMs can match or even surpass larger baselines, with our best configuration based on Qwen3-1.7B achieving an average score +9.2 points higher than Qwen3-32B. We release a comprehensive collection of all the publicly available Italian medical datasets for NLP tasks, together with our top-performing models. Furthermore, we release an Italian dataset of 126M words from the Emergency Department of an Italian Hospital, and 175M words from various sources that we used for continual pre-training.

Case Report Forms (CRFs) are structured instruments widely used in clinical research to systematically collect patient information according to predefined protocols. In practice, CRFs are often manually completed by clinicians based on patients’ clinical reports, a process that is time-consuming and prone to inconsistencies. Despite their central role in medical studies, automatic population of CRFs from clinical narratives remains largely underexplored in the Natural Language Processing community, partly due to the scarcity of publicly available datasets. In this paper, we present the CRF Filling Shared Task, organized at the CL4Health Workshop at LREC 2026, which aims to advance research on automatic extraction of structured clinical information from unstructured patient notes. The task consists of assigning the correct value to a set of predefined CRF items given a clinical note. The target dataset is derived from a real-world CRF for dyspnea assessment, comprising 134 medical items with predefined value sets. The task is provided in two languages, Italian and English. We describe the dataset, the task formulation, and the evaluation framework, and discuss the participating systems and their results. By introducing this shared task, we aim to stimulate research on clinically applicable NLP systems for structured data extraction in healthcare.

Understanding procedural skills from visual data is a key challenge in medical AI, especially for tasks that require reasoning over temporal sequences. We report on FBK-NLP’s participation at the ClinSkill QA 2026 shared task, which requires models to arrange shuffled key frames into a coherent sequence of clinical actions and provide explanations for the resulting order. We conduct a systematic study of prompting and reasoning strategies using an open and easily deployable vision-language model (VLM). The central finding of our study is that incorporating keypoint-based representations of people’s body parts substantially improves temporal reasoning behind frame ordering. Furthermore, we show that model performance is highly sensitive to prompt design and to seemingly minor factors such as filename ordering and the inclusion of domain information.

This paper investigates proactivity, a characteristic phenomenon of collaborative human-human interaction, where a participant in the dialogue offers the addressee some useful and not explicitly requested information. More precisely, a proactive behaviour is: (i) self-prompted and not simply reactive, that is, the speaker does not act merely in response to the requests the other participant has made; (ii) somehow effective for the achievement of the dialogue goal, since the speaker has a long-term, goal-directed behaviour that predicts future states and needs. Proactivity has been poorly investigated from a theoretical point of view, and there is a general need of empirical data for both quantitative and qualitative research. The paper provides an extensive analysis of proactivity in several human-human task-oriented dialogic corpora, selected with different characteristics, including chat exchanges and telephone calls, collection modalities such as natural setting and Wizard of Oz, and two languages, Italian and English. The main result is the D-Pro Corpus, a new resource manually annotated at the utterance level with proactivity and dialogue acts, which allows to investigate proactivity in the context of task-oriented dialogues. There are several findings from our empirical investigation of proactivity: (i) we find that about 20% of turns in our corpus are proactive turns, showing that this is a very diffused and relevant phenomenon; (ii) we confirm the non-reactive nature of proactivity, highlighting the presence of a pattern where a turn in the dialogue triggers a reaction in a following turn and a proactive utterance is then added to the turn; (iii) we show that only a limited number of dialogue acts are actually involved in expressing proactivity, and we discuss the theoretical implications of this finding; (iv) we empirically confirm that proactivity has a crucial role in recovering from goal-failure situations, contributing to the effectiveness of the whole dialogue; (v) we support the intuition of a non-uniform distribution of proactive utterances throughout the dialogue. Our empirical findings and the D-Pro Corpus provide relevant insights for deeper theoretical investigations, as well as crucial resources for improving proactivity in current task-oriented dialogue systems.