Research Papers

True intelligence hinges on the ability to uncover and leverage hidden causal relations. Despite significant progress in AI and computer vision (CV), there remains a lack of benchmarks for assessing models' abilities to infer latent causality from complex visual data. In this paper, we introduce \textsc{\textbf{Causal3D}}, a novel and comprehensive benchmark that integrates structured data (tables) with corresponding visual representations (images) to evaluate causal reasoning. Designed within a systematic framework, Causal3D comprises 19 3D-scene datasets capturing diverse causal relations, views, and backgrounds, enabling evaluations across scenes of varying complexity. We assess multiple state-of-the-art methods, including classical causal discovery, causal representation learning, and large/vision-language models (LLMs/VLMs). Our experiments show that as causal structures grow more complex without prior knowledge, performance declines significantly, highlighting the challenges even advanced methods face in complex causal scenarios. Causal3D serves as a vital resource for advancing causal reasoning in CV and fostering trustworthy AI in critical domains.

Explainable AI (XAI) holds significant promise for enhancing the transparency and trustworthiness of AI-driven threat detection in Security Operations Centers (SOCs). However, identifying the appropriate level and format of explanation, particularly in environments that demand rapid decision-making under high-stakes conditions, remains a complex and underexplored challenge. To address this gap, we conducted a three-month mixed-methods study combining an online survey (N1=248) with in-depth interviews (N2=24) to examine (1) how SOC analysts conceptualize AI-generated explanations and (2) which types of explanations are perceived as actionable and trustworthy across different analyst roles. Our findings reveal that participants were consistently willing to accept XAI outputs, even in cases of lower predictive accuracy, when explanations were perceived as relevant and evidence-backed. Analysts repeatedly emphasized the importance of understanding the rationale behind AI decisions, expressing a strong preference for contextual depth over a mere presentation of outcomes on dashboards. Building on these insights, this study re-evaluates current explanation methods within security contexts and demonstrates that role-aware, context-rich XAI designs aligned with SOC workflows can substantially improve practical utility. Such tailored explainability enhances analyst comprehension, increases triage efficiency, and supports more confident responses to evolving threats.

The rapid proliferation of Large Language Models (LLMs) has raised significant trustworthiness and ethical concerns. Despite the widespread adoption of LLMs across domains, there is still no clear consensus on how to define and operationalise trustworthiness. This study aims to bridge the gap between theoretical discussion and practical implementation by analysing research trends, definitions of trustworthiness, and practical techniques. We conducted a bibliometric mapping analysis of 2,006 publications from Web of Science (2019-2025) using the Bibliometrix, and manually reviewed 68 papers. We found a shift from traditional AI ethics discussion to LLM trustworthiness frameworks. We identified 18 different definitions of trust/trustworthiness, with transparency, explainability and reliability emerging as the most common dimensions. We identified 20 strategies to enhance LLM trustworthiness, with fine-tuning and retrieval-augmented generation (RAG) being the most prominent. Most of the strategies are developer-driven and applied during the post-training phase. Several authors propose fragmented terminologies rather than unified frameworks, leading to the risks of "ethics washing," where ethical discourse is adopted without a genuine regulatory commitment. Our findings highlight: persistent gaps between theoretical taxonomies and practical implementation, the crucial role of the developer in operationalising trust, and call for standardised frameworks and stronger regulatory measures to enable trustworthy and ethical deployment of LLMs.

As machine learning models become increasingly prevalent in time series applications, Explainable Artificial Intelligence (XAI) methods are essential for understanding their predictions. Within XAI, feature attribution methods aim to identify which input features contribute the most to a model's prediction, with their evaluation typically relying on perturbation-based metrics. Through systematic empirical analysis across multiple datasets, model architectures, and perturbation strategies, we reveal previously overlooked class-dependent effects in these metrics: they show varying effectiveness across classes, achieving strong results for some while remaining less sensitive to others. In particular, we find that the most effective perturbation strategies often demonstrate the most pronounced class differences. Our analysis suggests that these effects arise from the learned biases of classifiers, indicating that perturbation-based evaluation may reflect specific model behaviors rather than intrinsic attribution quality. We propose an evaluation framework with a class-aware penalty term to help assess and account for these effects in evaluating feature attributions, offering particular value for class-imbalanced datasets. Although our analysis focuses on time series classification, these class-dependent effects likely extend to other structured data domains where perturbation-based evaluation is common.

As part of the Frontier AI Safety Commitments agreed to at the 2024 AI Seoul Summit, many AI developers agreed to publish a safety framework outlining how they will manage potential severe risks associated with their systems. This paper summarises current thinking from companies, governments, and researchers on how to write an effective safety framework. We outline three core areas of a safety framework - risk identification and assessment, risk mitigation, and governance - and identify emerging practices within each area. As safety frameworks are novel and rapidly developing, we hope that this paper can serve both as an overview of work to date and as a starting point for further discussion and innovation.

In 2024, the outbreak of Human Metapneumovirus (HMPV) in China, which later spread to the UK and other countries, raised significant public concern. While HMPV typically causes mild symptoms, its effects on vulnerable individuals prompted health authorities to emphasize preventive measures. This paper explores how sentiment analysis can enhance our understanding of public reactions to HMPV by analyzing social media data. We apply transformer models, particularly XLNet, achieving 93.50% accuracy in sentiment classification. Additionally, we use explainable AI (XAI) through SHAP to improve model transparency.

To address the critical need for secure IoT networks, this study presents a scalable and lightweight curriculum learning framework enhanced with Explainable AI (XAI) techniques, including LIME, to ensure transparency and adaptability. The proposed model employs novel neural network architecture utilized at every stage of Curriculum Learning to efficiently capture and focus on both short- and long-term temporal dependencies, improve learning stability, and enhance accuracy while remaining lightweight and robust against noise in sequential IoT data. Robustness is achieved through staged learning, where the model iteratively refines itself by removing low-relevance features and optimizing performance. The workflow includes edge-optimized quantization and pruning to ensure portability that could easily be deployed in the edge-IoT devices. An ensemble model incorporating Random Forest, XGBoost, and the staged learning base further enhances generalization. Experimental results demonstrate 98% accuracy on CIC-IoV-2024 and CIC-APT-IIoT-2024 datasets and 97% on EDGE-IIoT, establishing this framework as a robust, transparent, and high-performance solution for IoT network security.

Probabilistic Explanations for Linear Models

Bernardo Subercaseaux, Marcelo Arenas, Kuldeep S Meel

Formal XAI is an emerging field that focuses on providing explanations with mathematical guarantees for the decisions made by machine learning models. A significant amount of work in this area is centered on the computation of "sufficient reasons". Given a model $M$ and an input instance $\vec{x}$, a sufficient reason for the decision $M(\vec{x})$ is a subset $S$ of the features of $\vec{x}$ such that for any instance $\vec{z}$ that has the same values as $\vec{x}$ for every feature in $S$, it holds that $M(\vec{x}) = M(\vec{z})$. Intuitively, this means that the features in $S$ are sufficient to fully justify the classification of $\vec{x}$ by $M$. For sufficient reasons to be useful in practice, they should be as small as possible, and a natural way to reduce the size of sufficient reasons is to consider a probabilistic relaxation; the probability of $M(\vec{x}) = M(\vec{z})$ must be at least some value $δ\in (0,1]$, for a random instance $\vec{z}$ that coincides with $\vec{x}$ on the features in $S$. Computing small $δ$-sufficient reasons ($δ$-SRs) is known to be a theoretically hard problem; even over decision trees--traditionally deemed simple and interpretable models--strong inapproximability results make the efficient computation of small $δ$-SRs unlikely. We propose the notion of $(δ, ε)$-SR, a simple relaxation of $δ$-SRs, and show that this kind of explanation can be computed efficiently over linear models.

Instruction fine-tuning of large language models (LLMs) is a powerful method for improving task-specific performance, but it can inadvertently lead to a phenomenon where models generate harmful responses when faced with malicious prompts. In this paper, we explore Low-Rank Adapter Fusion (LoRA) as a means to mitigate these risks while preserving the model's ability to handle diverse instructions effectively. Through an extensive comparative analysis against established baselines using recognized benchmark datasets, we demonstrate a 42\% reduction in the harmfulness rate by leveraging LoRA fusion between a task adapter and a safety adapter, the latter of which is specifically trained on our safety dataset. However, we also observe exaggerated safety behaviour, where the model rejects safe prompts that closely resemble unsafe ones

As AI systems increasingly make critical decisions, deceptive AI poses a significant challenge to trust and safety. We present Self-Other Overlap (SOO) fine-tuning, a promising approach in AI Safety that could substantially improve our ability to build honest artificial intelligence. Inspired by cognitive neuroscience research on empathy, SOO aims to align how AI models represent themselves and others. Our experiments on LLMs with 7B, 27B, and 78B parameters demonstrate SOO's efficacy: deceptive responses of Mistral-7B-Instruct-v0.2 dropped from 73.6% to 17.2% with no observed reduction in general task performance, while in Gemma-2-27b-it and CalmeRys-78B-Orpo-v0.1 deceptive responses were reduced from 100% to 9.3% and 2.7%, respectively, with a small impact on capabilities. In reinforcement learning scenarios, SOO-trained agents showed significantly reduced deceptive behavior. SOO's focus on contrastive self and other-referencing observations offers strong potential for generalization across AI architectures. While current applications focus on language models and simple RL environments, SOO could pave the way for more trustworthy AI in broader domains. Ethical implications and long-term effects warrant further investigation, but SOO represents a significant step forward in AI safety research.

Accurately describing images with text is a foundation of explainable AI. Vision-Language Models (VLMs) like CLIP have recently addressed this by aligning images and texts in a shared embedding space, expressing semantic similarities between vision and language embeddings. VLM classification can be improved with descriptions generated by Large Language Models (LLMs). However, it is difficult to determine the contribution of actual description semantics, as the performance gain may also stem from a semantic-agnostic ensembling effect, where multiple modified text prompts act as a noisy test-time augmentation for the original one. We propose an alternative evaluation scenario to decide if a performance boost of LLM-generated descriptions is caused by such a noise augmentation effect or rather by genuine description semantics. The proposed scenario avoids noisy test-time augmentation and ensures that genuine, distinctive descriptions cause the performance boost. Furthermore, we propose a training-free method for selecting discriminative descriptions that work independently of classname-ensembling effects. Our approach identifies descriptions that effectively differentiate classes within a local CLIP label neighborhood, improving classification accuracy across seven datasets. Additionally, we provide insights into the explainability of description-based image classification with VLMs.

AI evaluations are an important component of the AI governance toolkit, underlying current approaches to safety cases for preventing catastrophic risks. Our paper examines what these evaluations can and cannot tell us. Evaluations can establish lower bounds on AI capabilities and assess certain misuse risks given sufficient effort from evaluators. Unfortunately, evaluations face fundamental limitations that cannot be overcome within the current paradigm. These include an inability to establish upper bounds on capabilities, reliably forecast future model capabilities, or robustly assess risks from autonomous AI systems. This means that while evaluations are valuable tools, we should not rely on them as our main way of ensuring AI systems are safe. We conclude with recommendations for incremental improvements to frontier AI safety, while acknowledging these fundamental limitations remain unsolved.