Large Language Models (LLMs) such as GPT-4 and LLaMA have demonstrated promising performance in fact-checking tasks, particularly in labeling the veracity of claims. However, the real-world utility of such fact-checking systems depends not only on label accuracy but also on the faithfulness of the justifications they provide. Prior work has explored various prompting strategies to elicit reasoning from LLMs, but most studies evaluate these styles in isolation or focus solely on veracity classification, neglecting the impact on explanation quality. This study addresses that gap by investigating how different prompt styles affect both the accuracy and the faithfulness of LLM-generated claim labelling and justifications. Seven established prompting strategies such as Chain-of-Thought, Role-Based, or Decompose-and-Verify, were tested across two datasets (QuanTemp and HoVer) using two efficient models: LLaMA 3.1:8B and GPT-4o-mini. Additionally, two novel prompt variants were introduced and all styles were tested under three label conditions to assess bias and explanation drift.

Large Language Models (LLMs) are increasingly used to generate fact-checking explanations, but evaluating how faithful these justifications are remains a major challenge. In this paper, we examine how well four popular automatic metrics—G-Eval, UniEval, FactCC, and QAGs—capture faithfulness compared to expert-written explanations. We look at how these metrics agree with each other, how they correlate with explanation similarity, and how they respond to controlled errors.Our findings show that while some metrics like UniEval and FactCC show some sensitivity to noise and partial alignment with expert reasoning, none of them reliably catch hallucinations or consistently reflect true faithfulness. Manual analysis also reveals that metric behavior varies depending on the type and structure of the claim. Overall, current metrics are only moderately effective and often biased toward the style of LLM-generated text. This study points to the need for more reliable, context-aware evaluation methods and offers practical insights for improving how we measure faithfulness in fact-checking tasks.

Large Language Models (LLMs) are becoming more commonplace in today's society. However their adoption rate, especially in the fact checking field, is being slowed down by the distrust in their thinking process and the rationales leading to the results. In crucial moments the justifications behind a verdict are more important than the verdict itself. However, LLMs often produce explanations that are not grounded in the provided evidence, leading to hallucinations and reduced trust in their outputs. This paper aims to show exactly the level the LLMs have reached in both the faithfulness of their explanations, based on some provided facts, and the correctness of their explanations. To investigate this, multiple LLMs are asked to assign a label to a claim based on some evidence provided from two datasets of varying complexity: HoVer and QuanTemp. The outputs are then evaluated both manually and by another LLM to evaluate how well the LLM relates to the evidence and if the LLM hallucinates in some parts of its responses. The results reveal that while some models demonstrate high correctness in label assignment, faithfulness in explanations varies significantly across models and evidence types. The outcomes of this experiment aim to inform both LLM developers and fact-checking researchers about the current limitations of LLMs in response quality while also showing which areas require further improvements to become mainstream.

Rising mental health issues among adolescents have increased interest in automated approaches for detecting early signs of psychological distress in digital text. One important focus is the identification of cognitive distortions – irrational thought patterns – because of their role in aggravating mental distress, and early detection may enable timely, low cost interventions. While prior work has focused on English data, we present a first in-depth study of cross lingual and cross register generalization for cognitive distortion detection, using forum posts written by Dutch adolescents. We frame the task at two levels: (1) detecting whether a post contains a cognitive distortion, and (2) identifying the specific text span that expresses it. Our findings show that domain adaptation methods perform best for post-level detection, while a simpler technique – sentence embeddings with a classifier – outperforms more complex models for span identification. Results show predicting cognitive distortions in text is challenging, and highlight how changes in language and writing style can significantly impact performance.

Suicide is a leading cause of death, yet predicting it remains a significant challenge. Risk factors such as depression or substance use are commonly used for prediction, but their predictive performance is often only slightly better than chance. Additionally, many cases go undetected due to a lack of contact with mental health services. Social media, however, offers a unique opportunity, as people often share their thoughts and struggles online in real time. In this work, we propose a novel task and method to approach it: predicting suicidal ideation and behavior (SIB) before a user ever expresses it on an online forum. This predictive framing, where no self-disclosure is used as input at any stage, remains largely unexplored in the suicide prediction literature. Our model, Early-SIB, achieves a balanced accuracy of 0.73 for predicting future SIB on a Dutch youth forum, demonstrating that such tools can offer a meaningful addition to traditional methods.

Large Language Models (LLMs) are increasingly transforming how scientists approach research, with emerging tools supporting ideation, experimentation, and publication in attempts to expedite the research process. This work focuses on the foundational first step: generating novel, high-quality research ideas. Building on an existing LLM idea generation method, we introduce a system for personalised research idea generation that generates ideas by analysing a researcher’s publication history, providing tailored research recommendations. We evaluate the generated ideas via expert surveys across predefined metrics, offering insights into the system’s effectiveness and LLMs’ role in ideation in general. Findings show that personalised generation produces novel, interesting, and potentially impactful ideas comparable to baseline methods. Clarity remains a common limitation, while feasibility improves in the personalised variant, highlighting the potential of personalised LLM ideation.

The use of research assistants has increased significantly, providing support and automation for researchers. However, there is limited research on researchers using research assistants and what assistance researchers require for each research stage.
We interview researchers to gather insights into their opinions and usage, and afterwards develop a prototype. Researchers highlight that the most difficult part of research is the experiment design, which is reflected in the lack of literature.
Participants use research assistants for reading literature and writing, but request more support. The research assistant must be factual, transparent, and correct, and including a conversation allows for feedback and discussion.
We evaluate the prototype for the experiment design phase, highlighting the effectiveness of the component architecture by generating correct experiments.

We investigate the application of Retrieval-Augmented Generation (RAG) for enhancing the analysis of corporate sustainability disclosures. We introduce CorSus, a novel dataset for evaluating RAG models in answering corporate sustainability-focused claims, using data from the Transition Pathway Initiative for over 100 companies. Further, we develop a subset of this dataset with reference documentation and fully explained answers. Finally, in a systematic framework, we optimise and benchmark state-of-the-art RAG approaches using the CorSus dataset. With this work, we aim to empower stakeholders with a tool for informed evaluations of corporate sustainability practices, thereby encouraging a greater commitment to environmental responsibility.

With the increasing popularity of Large Language Models (LLMs), fine-tuning them has become increasingly computationally expensive. Parameter Efficient Fine-Tuning (PEFT) methods like LoRA and Adapters, introduced by Microsoft and Google, respectively, aim to reduce the number of trainable parameters, with the current state-of-the-art combining both methods as LoRA Adapters. This paper introduces Transformer Modules as a PEFT method. These modules utilize Modular Transformer Blocks (MTBs) inserted into a frozen pre-trained model, achieving competitive performance while significantly reducing computation costs. Compared to the current state-of-the-art using GPT-2, BERT, and T5, Transformer Modules further reduced compute time by 39.7\% and training memory by 72.7\%, with a performance cost of 4.5±2.51\% on the GLUE benchmark. Additionally, the paper presents the Transformer Bridge, a continuous vector transformer designed to transfer Transformer Modules across different models. This could enable cross-model fine-tuning, allowing model-agnostic modules, such as an ethics or medical module, to be used across various LLMs without retraining or access to the original dataset. Although the current implementation of the Transformer Bridge did not fully succeed in mapping embedding spaces, analysis of the results suggests that further refinements using traditional model distillation techniques could lead to success in future iterations.

Water management systems (WMSs) are complex systems in which often multiple conflicting objectives are at stake. Reinforcement Learning (RL), where an agent learns through punishments and rewards, can find trade-offs between these objectives. This research studies three case studies of WMS simulations in the context of RL problems and notes their similarities and differences. Based on these, core properties of WMSs are defined and used to formulate a general WMS as a RL problem. This bottom-up approach uses Gymnasium to implement the RL problem. The result is compared to a simulation from one of the case studies and produces the same results. While maintaining this level of accuracy, it is applicable to a much wider range of WMSs. It thereby contributes to generalisation of WMSs in the context of RL, and removes the need to rewrite simulations each time.

Efficient management of water resources is increasingly critical in the face of growing challenges such as climate change and population growth. This research paper introduces RL4Water, an adaptable framework for simulating water management systems using multi-objective reinforcement learning (MORL). Adhering to the Gymnasium API standard, RL4Water ensures seamless integration with existing MORL algorithms. The framework includes diverse facility classes to accurately model the physical components of water networks. Its generalizability is enhanced by allowing users to modify both the physical properties of these components and the key features of the MORL simulations. RL4Water's capabilities are demonstrated through two case studies: simulations of the Nile River and the Susquehanna River, validating its accuracy and flexibility in managing both large, distributed water systems and centralized systems with complex reservoirs. By bridging the gap between water management and reinforcement learning, RL4Water offers a unified platform for developing and researching water management simulations.

This study investigates the use of Multi-Objective Natural Evolution Strategies (MONES) to optimise water management control policies in the Nile River Basin, focusing on four key objectives: minimising irrigation deficits for Egypt and Sudan, maximising hydropower production for Ethiopia, and maintaining water level in the High Aswan Dam (HAD). The developed Nile River simulation was integrated with MONES and used to train an agent for making release decisions. Performance metrics including hypervolume, epsilon-indicator, and Inverted Generational Distance Plus (IGD+) were employed to compare MONES with the EMODPS baseline. The results indicate that while MONES identifies feasible solutions, it falls short in exploration and overall performance compared to EMODPS. The study contributes to the development of water management strategies by open-sourcing Nile River framework compatible with reinforcement learning and demonstrating the potential and limitations of MONES in multi-objective optimisations.

This paper studies the simulation of the Nile River as a multi-objective reinforcement learning problem. The main goal of this essay is to develop and evaluate the visualization techniques to effectively present the results of reinforcement learning models. Using a multi-objective approach, visualizations are very important for understanding the trade-offs and complexities in managing the Nile River problem.

This study includes a user evaluation to compare different visualizations, analyzing their effectiveness in terms of their clarity and usefulness using ANOVA test. Additionally, the effectiveness of clustering and full data points will be analyzed using a chi-square test to choose which visualisation technique works the best.

According to the results, stacked bar chart and parallel coordinates plot performed the best, while the spider plot performed the worst. Additionally, there is no preference between clustered and full data points visualizations based on the user evaluation.

This paper explores the application of evolutionary algorithms to enhance task generation for Neural Processes (NPs) in meta-learning. Meta-learning aims to develop models capable of rapid adaptation to new tasks with minimal data, a necessity in fields where data collection is costly or difficult. By integrating evolutionary strategies, we aim to enhance the efficiency and robustness of NPs. We evaluate our approach using 1-D function regression problems, where Genetic Algorithm generates diverse and challenging tasks. Our results show that the evolutionary approach improves learning efficiency and model performance, achieving lower Root Mean Squared Error (RMSE) compared to traditional methods.